The Future of Retrosynthesis and Synthetic Planning: Algorithmic, Humanistic Or the Interplay?

CSIRO PUBLISHING Aust. J. Chem. 2021, 74, 291–326 Cornforth Review https://doi.org/10.1071/CH20371

The Future of Retrosynthesis and Synthetic Planning: Algorithmic, Humanistic or the Interplay?

Craig M. Williams A,B and Madeleine A. Dallaston A

ASchool of Chemistry and Molecular Biosciences, University of Queensland, Brisbane, Qld 4072, Australia. BCorresponding author. Email: [email protected]

The practice of deploying and teaching retrosynthesis is on the cusp of considerable change, which in turn forces practitioners and educators to contemplate whether this impending change will advance or erode the efficiency and elegance of organic synthesis in the future. A short treatise is presented herein that covers the concept of retrosynthesis, along with exemplified methods and theories, and an attempt to comprehend the impact of artificial intelligence in an era when freely and commercially available retrosynthetic and forward synthesis planning programs are increasingly prevalent. Will the computer ever compete with human retrosynthetic design and the art of organic synthesis?

Keywords: retrosynthesis, biomimetic synthesis, natural products, computer assisted synthesis, total synthesis, computer assisted retrosynthesis, artificial intelligence, antithetic analysis.

Received 21 December 2020, accepted 25 March 2021, published online 12 May 2021

Introduction synthesis’.[2] It should be recognised, however, that many tal- Retrosynthesis (or retrosynthetic analysis or antithetic analysis) ented and famous scientists performing organic synthesis were is defined as ‘a problem-solving technique for transforming the already utilising some form of innate retrosynthesis before the structure of a synthetic target molecule to a sequence of pro- term was coined. From an educational perspective, Warren gressively simpler structures along a pathway which ultimately masterfully described synthetic planning using the disconnec- leads to simple or commercially available starting materials for a tion approach,[3] which is an intuitive method that likely chemical synthesis’.[1] This definition was formulated and embodied early thinking. Overall, however, it is hard to ascer- underpinned by extensive research endeavours in heuristic tain whether the very early methods preceding Corey’s anti- chemical synthesis performed by the research group led by Elias thetic analysis focussed on bond breaking (disconnection), or James Corey, who won the 1990 Nobel Prize in Chemistry ‘for more on literature-inspired bond formation (i.e. literature his development of the theory and methodology of organic knowledge of chemical reaction methodology), or both,[4] but in

Craig M. Williams (C.M.W.) was born in Adelaide, Australia. He received his B.Sc. (Hons) degree in chemistry in 1994 from Flinders University. In 1997, he was awarded his Ph.D. in organic chemistry from the same institution under the supervision of Professor Rolf H. Prager. He undertook post-doctoral studies as an Alexander von Humboldt Fellow working with Professor Armin de Meijere at the Georg-August-Universita¨t, Go¨ttingen, Germany, from 1997 to 1999. In early 1999, he accepted a second post-doctoral fellowship at the Australian National University with Professor Lewis N. Mander. Professor Williams has held an academic position at the University of Queensland since 2000, and during this time has won a number of awards including a Thieme Chemistry Journals Award in 2007, an Australian Research Council Future Fellowship award in 2011, and the Award for Outstanding Contribution to Research (SCMB, UQ, 2019). The Williams research group explores numerous interests within the discipline of organic chemistry (e.g. medicinal chemistry, fundamental molecules, natural product isolation, microelectronics, drug and agrichemical development, impact sensitive molecules) enabled by organic synthesis refined through the construction of biologically active complex natural products (diterpenes, polyketides, alkaloids), and designs synthetic methodology to assist in this endeavour (synthetic transformations and reagents). Professor Williams especially enjoys teaching whole molecule retrosynthesis to undergraduate and post-graduate students. Madeleine A. Dallaston (M.A.D.) was born in Leicester, England, and grew up in Brisbane, Australia. In 2016, she received her B.Sc. from Griffith University, earning the RACI Queensland Branch Prize for that year. In 2017, she undertook a B.Sc. (Hons) at the University of Queensland in collaboration with Defense Science and Technology, working on energetic materials for countermeasures development. She then commenced her Ph.D. in organic chemistry at the University of Queensland under the supervision of Professor Williams and G. Paul Savage in 2018, focusing on the synthesis of novel heterocycles for application as bioisosteres.

Journal compilation Ó CSIRO 2021 Open Access CC BY www.publish.csiro.au/journals/ajc 292 C. M. Williams and M. A. Dallaston modern times, the situation is rapidly changing. For example, reviews covering synthetic achievements by subject (e.g. drug 2(1 + 5 + 5) + (5 + 3) = 30 synthesis[5] and natural product synthesis[6–8]) are generating highly refined retrosynthetic approaches, as are reaction popu- larity assessments (e.g. medicinal chemistry[9]) and modern methodology-focussed retrosynthetic planning (e.g. C–H activation[10] and photochemistry[11]). The most evolutionary change currently being confronted by the synthetic community, 1 however, is the realisation that computer-aided retrosynthesis,[12] and the accompanying artificial intelligence (AI) [13] that proposes forward syntheses, may very soon become O CI commonly available to educators and practitioners. It is this Li + verge of revolutionary retrosynthesis that has inspired the 2 CI O present review, and as such the authors will attempt to shed light 3 on a subject that has been in motion since the late 1960s, but not FGI in the past widely adopted. To assist in achieving the goal of this FGI perspective, classical and historical methods and practices of Br retrosynthesis will be briefly overviewed (i.e. using examples OH and experiences well known to the authors), along with an 4 5 assessment of currently available retrosynthesis programs. FGI Combined, this assessment will hopefully generate philosophical debate around the pros and cons of AI introduction and FGI contemplation on the future of organic synthesis design. OH CN 7 6 O Retrosynthetic Methods and Practices Scheme 1. Cornforth’s synthesis of squalene (1) using a high proportion of There are essentially two forms of retrosynthesis, and these are FGIs. Retrosynthesis arrows introduced by Corey and reinforced by Warren highly integrated. The first is functional group interconversion are used throughout.[16] (FGI) or functional group exchanges, and the second is whole- molecule (or target-orientated) retrosynthesis. FGI plays a role in both, whereby the first mode is the ability fragrances, pharmaceuticals, agrichemicals, materials, natural to navigate through a range of functional groups in a linear products, and environmental and fundamental molecules. fashion to arrive at an alternative functional group (this also With all these drivers and inspiration to make such mole- includes protecting groups). An example to highlight this situa- cules, members of the synthetic community have resorted to numerous retrosynthetic approaches to achieve successful com- tion is the stereoselective synthesis of squalene (1) in 1959, [17a] which was performed by husband-and-wife team Drs John and pletion of their proposed targets (see below). Interestingly, Rita Cornforth and, presumably, their joint student K. the majority of these approaches were developed through the Mathew.[14] At this time, the term retrosynthesis did not exist; process of performing natural product total synthesis, mainly because natural products were the first on the scene during the however, conceptual thinking around construction of the target [17b] was discussed. Interestingly, the approach presented focussed emergence of the discipline of organic chemistry. Following on the number of carbon subunits required to achieve the 30- isolation and subsequent elucidation of a natural product, there carbon framework (i.e. in terms of numerical values entered into was an immediate structural challenge, often augmented by an equation[15]), and that a reported olefin synthesis method intriguing biology activity, that provided a substantial synthetic would be used. This is a clear example of deploying a literature- challenge. This is not to say that non-natural targets have not inspired bond formation strategy, and to achieve this goal a large contributed to the art, but the fact that Mother Nature has for number of FGIs were undertaken (Scheme 1). To exemplify this eternity been generating a vast library of molecules has pre- point, homogeranyl lithium (2) was required to add to dichloro- sented a ‘spoiled for choice’ situation for synthetic enthusiasts diketone (3) in the final stages of the synthesis, and this was and educationalists. derived from homogeranyl bromide (4). Bromide 4 was in turn The various retrosynthetic approach concepts are outlined obtained from alcohol 5, which was prepared via reduction of below using select examples well known to the authors. homogeranic acid (6). Finally, base hydrolysis of geranyl cya- nide (7) afforded 6 (Scheme 1). Overall, five FGIs were Biomimetic Synthesis deployed to transform 7 into 2, but the homogeranyl carbon Biomimetic, or biogenetically inspired, synthesis continues to be framework went unchanged, i.e. only the end functional group is a regularly deployed method to construct complex natural pro- different. ducts. In essence, this area evolved from natural product isolation Although arguably FGI can also be considered target- and synthetic chemists postulating how organisms might pro- orientated synthesis, the latter term is really reserved for targets duce secondary metabolites through biosynthesis and/or that require as part of their retrosynthetic planning a combina- accompanying metabolic processes. The retrosynthetic planning tion of carbon framework manipulations and FGIs. Examples process therefore adopts synthetic transformations, and/or key that fit this category include small, large, simple, and complex intermediates, that mimic a postulated biological process (e.g. molecular targets, and thus are only limited by academic enzymatic oxidation) undertaken by the organism.[18] It should imagination and industrial desires. Some (of the many) areas be noted that domino,[19] tandem,[20] and cascade[21] reactions that have benefitted from these endeavours include flavours, are prevalent, but importantly inspired by nature. The Future of Retrosynthesis 293

An example that embraces both biomimetic and domino- of natural products. Equally, it could be considered that total based retrosynthetic planning is the total synthesis of diterpenes synthesis of vibsanin B (8) might be a viable starting point, but in the vibsane series of natural products.[22] Interestingly, the approaches to construct the 11-membered ring are daunting in isolation of this diterpene class was inspired by reports that the comparison with developing strategies for decorating a six- scrub Viburnum awabuki (Caplifoliaceae) was used as a fish membered ring of the type seen in 9. With these considerations poison by Japanese villagers (i.e. piscicidal activity).[23] Since front of mind, the following retrosynthesis was formulated that time, other viburnum species (i.e. V. odoratissimum and (Scheme 3). It was opted to install the enol ester side chain last, V. suspensum) have yielded a wide variety of different structure giving rise to 17, which although carrying the requisite retrons types all proposed to be biosynthetically related, and many was better equated as 18 to hold oxygenation at the doubly displaying pronounced neurite outgrowth activity.[24] Of rele- allylic position. Access to 18 would likely arise from a combi- vance to this section are the neovibsanins, which Fukuyama[25] nation of alkylation and addition to cyclohexenone 19, with proposed were biosynthetically derived from vibsanin B (8), via the overall sequence starting from methylcyclohexenone 20 key intermediate 9 (Scheme 2). Intermediate 9 was then pro- (Scheme 3). posed to undergo a series of cyclisations and rearrangements, the Although commercially available at the time, it was more first being tetrahydrofuran formation to give 10, which is poised convenient to synthesise methylcyclohexenone 20 in-house on to undergo further cyclisation to neovibsanin A (11, via path A), large scale from ethyl acetylacetate (21).[26] Addition of homo- or provide carbocation 12 via path B. Carbocation 12 can then be prenyl magnesium bromide (22)to20 could be achieved using attacked either intermolecularly by water to give neovibsanin H standard cuprate conditions on reasonable scale in the racemic (13, via path C), or intramolecularly by the pendant homo- series (i.e. 23),[27] but also asymmetrically using the chiral isoprenyl alkene to give the bicyclic cation 14 (via path D). ligand 24, affording the 11R-isomer 25 (Scheme 4). Dehydro- Cation 14 is then open to either nucleophilic attack by water, genation using IBXNMO (26; IBX ¼ o-iodoxybenzoic acid, resulting in the formation of neovibsanin F (15, via path E), NMO ¼ N-methylmorpholine N-oxide)[28] gave 27, setting the or elimination via path F to afford neovibsanin G (16) stage for introduction of the important methylene hydroxy group (Scheme 2). (i.e. 28) via a Morita–Baylis–Hillman reaction. Ultimately, this Digestion of this extensive biosynthetic proposal suggested was achieved by developing a green chemistry method using the that if a candidate similar to intermediate 9 was entertained common surfactant sodium dodecyl sulfate (SDS) in water, during retrosynthetic analysis, then via multiple acid-catalysed which could be performed at scale.[29] Silyl protection of the steps, it might be possible to access the entire neovibsanin class pendant hydroxyl group enabled enolate-mediated alkylation of ketone 28 with ethyl iodoacetate, which gave a mixture of diastereomers 29 and 30. The major diastereomer 30 was then MeO O [30] + reacted with the acyl anion equivalent, lithium dithiazide H O O H (31)[31] to give lactone 32. Unmasking the aldehyde was O O O O performed using mercury acetate, and the final methyl enone OH was installed using ylid 33, to afford 34 (Scheme 4).[32] O Neovibsanin A (11) Having access to an advanced intermediate (i.e. 34) that most Vibsanin B closely matched that proposed biosynthetically (i.e. intermedi- (8) Path A ate 9) enabled the next phase of the biomimetic synthesis to be + + H H investigated. An important aspect in moving forward in this O + O regard was the fact that many neovibsanin family members H H A A contained methoxy groups, presumably from methylation of the O HO B HO [33] O O corresponding hydroxyl group via plant methyl transferases. OH B Therefore, instead of attempting to emulate an aqueous O O

9 10 O O H Path B HO HO O O O O O OH OH O O O OH O O O Path D 2 OH + + 2 C 14 E 12 Esterification 9 17 F F H D Path F OH2 Path C Path E Neovibsanin G 1,2-Addition O (16) O O O O O O O OH O O O O OH O OH 20 OH 19 a-alkylation Neovibsanin F Neovibsanin H 18 (15) (13)

Scheme 2. Postulated biosynthetic pathways to the neovibsanins. Scheme 3. Retrosynthetic analysis of key intermediate 9. 294 C. M. Williams and M. A. Dallaston

O BrMg 22 11 PF6 Cu(OTf)2 N N + Et2O HO 25 24 70 % (91 % ee)

O O O BrMg 1) H2CO 22 OEt CuBr·SMe 21 HN / EtOH 2 20 THF D 2) H2SO4 /

O OH O O

H2CO / DMAP IBX·NMO

SDS / H2O DMSO 48 % 64 % (12 g scale) 28 27 23

1) TBSCI / Imid 2) LDA / ethyl iodoacetate

O N O S EtO S N Li O OTBS S 31 S O OTBS THF 57 % a = 29 (24 %) b = 30 (37 %) 32

O O 1) Hg(CIO4)2 O– O HO O I THF / H O N 2 + OTBS O O Ph P 2) 3 O 33 O 26 CHCI / D 34 3 66 % 2 steps

Scheme 4. Forward synthesis of key intermediate 34. environment or conditions, methanol could be used as both the However, it is worthwhile to note that even the most well- solvent and nucleophile. Accordingly, 34 was dissolved in considered retrosynthetic planning can come unstuck, and this methanol and treated with concentrated sulfuric acid at room occurred when attempting the total synthesis of neovibsanins A temperature overnight. The strongly acidic conditions promoted and B.[35] It was thought that the proposed advanced intermedi- several cascading reactions, the first of which was deprotection ate 34 was suitably flexible to follow the proposed biosynthetic of the t-butyldimethylsilyl (TBS) group to unmask the primary route, but one fault resided within 34, which amounted to a alcohol seen in 35. Oxy-Michael addition formed the tetrahy- single stereocentre that was opposite to that proposed for drofuran ring observed in 36, which underwent epimerisation biosynthetic intermediate 9 (Scheme 5, red highlight). On (i.e. 37), before solvolysis and attack by the solvent (path A) to treatment of 34 with concentrated sulfuric acid, although this give carboxylic acid 38. Under the reaction conditions, the acid time at 48C overnight (room temperature (r.t.) resulted in undergoes Fischer esterification to afford the methyl ester (39). decomposition), the resulting alcohol (43) did not undergo Global reduction followed by global oxidation provided alde- solvolysis, but instead acted as a nucleophile, giving rise to hyde 40, which was immediately subjected to conditions devel- ester 44. Ketalisation then afforded the 4,5-bis-epi-esters 45 and oped by Davies to install the elaborate enol ester sidechain. This 46, which were carried through via the respective aldehydes final manoeuvre delivered ()-2-O-methylneovibsanin H (41) (47 and 48) to the non-natural isomers 4,5-bis-epi-neovibsanin in 12 synthetic operations, whereby an acid-catalysed, one-pot, A(49) and B (50)(Scheme 5).[35] Of note is that the biological four-step sequence was key to achieving a concise total synthe- activity of 49 and 50 was very similar to that of neovibsanins A sis based on a biosynthetic proposal (Scheme 5).[32] and B.[35] Furthermore, a related outcome was experienced Based on these biogenetically inspired results, a similar when spirovibsanin A was targeted, in that ()-5,14-bis-epi- domino reaction, albeit combining Brønsted and Lewis acid spirovibsanin A (51) was obtained (Scheme 5).[36] promotion, successfully led to the asymmetric total synthesis Although retrosynthetic planning based on biosynthesis pos- of (–)-neovibsanin G (42) starting from 25 (Scheme 5).[34] tulates rarely utilises enzymatic methods for performing forward The Future of Retrosynthesis 295

O O O O O O O MeO C O O 2 1) LiAlH4 / THF 2 (±)-2-O-Methylneovibsanin H (41) O DMAP, µW O OMe OMe 2) DMP / CH2Cl2 110°C OMe 10 % over 3 steps Toluene 40 39

Sigma bond rotation Epimerisation Esterification 72 % O O O O O MeOH O O Deprotection B O O O O Michael O O O O HO2C O MeOH Addition A Path A OTBS OH HOMe H 2SO4 A Solvolysis: Retro-Michael A OMe Allyl cation 34 35 36 37 38 Lactone ring opening Path B R1 H + O O O H R2 4 5 H H O O HO MeO2C Hemiacetal O Ketal HO HO O MeO C O formation 2 formation OH MeO 73 % O O 1 2 43 44 R = OMe, R = Me (45) 9 R1 = Me, R2 = OMe (46)

1) LiAlH / Et O O 4 2 54 % O 2) pyr / DMP / CH2Cl2 O O O O O O O R1 H R1 H O O O O R2 R2 O 2 O OMe O (±)-5,14-Bis-epi- DMAP, µW O (-)-Neovibsanin Spirovibsanin A O 110°C G (42) (51) Toluene 4,5-Bis-epi-Neovibsanin 47 and 48 A (49) and B (50)

Scheme 5. Synthesis of the neovibsanins 42 and 51 and of the bis-epi-non-natural derivatives 49 and 50. synthesis, such synthetic transformations mediated by enzymatic (53)) reported in 1964 by Philip Eaton, and coworker Thomas control have found wide application in all other areas of organic Cole.[41] The sequence was later modified by the same authors synthesis.[37] That said, they are very substrate-dependant.[38] for conversion into the Platonic solid cubane (54), a seminal Lastly, from a philosophical perspective, adopting a retrosyn- contribution to the field of strained hydrocarbons.[42] Although a thetic plan based on naturally occurring biosynthetic considera- retrosynthetic approach was not discussed, as was the conven- tions could be considered plagiarism from Mother Nature. A tion in that era, the clear inspiration for the route taken hinged on similar argument could be mounted for AI, which presents a the fact that cyclopentadienone (55) underwent regioselective similar philosophical dilemma (see further below). [4þ2]-dimerisation.[43] However, at that time the endo stereoselective nature of these reactions was not defined, but was Linear Synthesis thought to be endo as drawn by Hafner and Goliasch in 1961.[44] Linear (or consecutive) syntheses are the most common out- Interestingly, Eaton independently confirmed the endo assign- comes of retrosynthetic planning, because they are by design ment via the known hydrocarbon 56, but the route was not iterative, and thus cater for both small and large target mole- verified with either experimental procedures or cited.[41] Appar- cules. Like all synthetic approaches to a target molecule, linear ently, this was achieved by reduction of dicyclopentadien-1,8- syntheses have their place; however, generally they are avoided dione (57) to the diketone (58), followed by conversion to the from both an elegance perspective (e.g. ideal syntheses)[39] and bisthioketal (59), and final reduction via Raney nickel to endo- efficiency point of view (e.g. step economy),[40] i.e. conver- tetrahydrodicyclopentadiene (56). These key pieces of knowl- gence approaches are preferred when possible (see below). edge regarding the endo-stereochemistry, combined with A landmark contribution of tactical linear retrosynthesis reports that chlorinated cyclopentadienones also undergo deployment is that of the cubane skeleton (i.e. cubane-1,4- [4þ2]-dimerisation,[43,45] enabled Eaton to consider the follow- dicarboxylic acid (52) and the corresponding dimethyl ester ing approach. Gaining access to a suitably substituted 296 C. M. Williams and M. A. Dallaston

CO R 2 R = H (52) RO2C R = Me (53) O X O O X H R R R X H 55 (R = H) 54 63 O 62 X 61 O 60 (R = X)

O O S S HHH H Pd/C/H2 Ni Raney H H H H S S O 57 58 O 59 56

Scheme 6. Retrosynthetic analysis of the cubane skeleton. cyclopentadienone (60) would provide a dihalogenated endo- O O O Br dicyclopentadien-1,8-dione (61). Although, given the photo- NBS 2 Br chemically sensitive bridgehead bromide seen in 61, a daring CCI þ 4 pentane / CH2Cl2 regioselective photolysis was postulated to drive a [2 2]- Br 65 0–10°C Br Br cycloaddition to form the precursor cage (62), a subsequent 64 66 Et O double ring contraction was then envisaged to give the 1,4- 2 EtNH substituted cubane framework (i.e. 63), enabled by the Favorskii –20°C 2 rearrangement (Scheme 6). Br O hv O In the forward direction, the synthesis started by treating MeOH / HCl Br O cyclopentenone (64) with N-bromosuccinimide, which gave the Br then H O / D monobrominated material 65 en route to the tribromide 66. Br 2 O 68 Br Exposing 66 to ethylamine in diethyl ether at low temperature 69 O 67 generated 2-bromo-cyclopentadienone (67), which underwent KOH 40 % 3 steps immediate dimerisation to afford the diketone dimer (68)in 50 % aq. 40 % yield over three steps. Subsequent UV irradiation effected the [2þ2]-cycloaddition to give the cage precursor 69, which O O MeOH / H SO was treated with potassium hydroxide to mediate the double OH 2 4 OMe Favorskii, and reveal cubane-1,4-dicarboxylic acid (52). The HO Δ MeO corresponding dimethyl ester (53) could be readily obtained O 52 30 % 3 steps O 53 through classical Fischer esterification conditions (i.e. methanol and catalytic amounts of sulfuric acid). Although there are now 1) SOCl2 numerous methods to access cubane itself (i.e. 54), a more 2) DMAP / CHCl3 expedient route consists of double decarboxylation of the diacid hv / Δ [46] (52), using modified Barton conditions that use chloroform S as the H atom donor (Scheme 7).[47] 54 N ONa Given the original questions around endo-selectivity concerning the cyclopentadienone dimerisation, Eaton justified this stereochemical configuration on the grounds that ‘interactions O Br of like dipoles should be minimized in the geometry of the O O Br transition state [i.e. 70] for Diels–Alder dimerization’ Br (Scheme 7).[41] 70 Since Eaton’s reports, there has been a great deal of activity 67 68 Br O O in terms of both broadening the understanding of and developing Br synthetic methods for the cubane system.[48] A large-scale industrial process was developed for the synthesis of dimethyl Scheme 7. Synthesis of cubane and the corresponding 1,4-disubstituted [49] cubane-1,4-dicarboxylate (53), which was found not to be analogues. impact-sensitive, unlike the corresponding diacid (52).[50] Fur- thermore, cubane has more recently become known as a phenyl ring isostere, and has been deployed as a bioisostere in medicinal that can be rapidly assembled in a convergent manner. This [51] and agricultural chemistry. approach saves considerable synthetic steps (and or manipulations) compared with other methods, in particular Convergent Synthesis linear sequences (see above).[52] Convergent syntheses are based on a branched retrosynthetic The Australian rainforest polyketide EBC-23 (71), isolated pathway such that the individual branches lead to smaller from Cinnamomum laubatii (family Lauraceae) in northern fragments (synthons) that are more synthetically tractable, and Queensland, is an example of a three-component convergent The Future of Retrosynthesis 297 synthesis.[53] The inspiration to undertake this collaborative quantities, starting from commercial epi-chlorohydrin (86). industry project included encouraging biological activity (i.e. TMS-dithiane (87) was prepared by treating dithiane (88) with growth inhibitor activity of the androgen-independent prostate n-butyl lithium (nBuLi) and quenching the resulting anion with tumour cell line DU145), and the necessity to confirm both the trimethylsilyl chloride (TMSCl). Subsequent treatment of proposed relative and absolute stereochemical assignment. TMS-dithiane (87) with nBuLi provided the TMS-dithiane Ancillary drivers included undertaking a total synthesis cam- anion (89), which when reacted with epoxide 84, followed by paign under project management milestone conditions, and addition of epi-chlorohydrin (86), afforded the desired disubsti- being mindful that the synthesis should be designed such that tuted dithiane (90). Interestingly, unmasking the acyl anion it was process chemistry transferrable for potential large-scale equivalent (i.e. dithiane moiety seen in 90) by oxidative removal production. with mercury perchlorate not only provided the ketone function, The retrosynthetic approach for EBC-23 started by discon- but also mediated TMS deprotection, giving the hydroxyketone nection of the EE-configured spiroketal unit, because it was 91. The unstable hydroxyketone (91) was immediately sub- plausible that this was occurring biosynthetically (i.e. selective jected to a Prasad syn-selective reduction using diethylmethox- oxidation of the polyol chain to a single keto moiety facilitating yborane and sodium borohydride,[58] and the resulting diol thermodynamic cyclisation via the anomeric effect), and it was protected as an acetonide (i.e. 92). All transformations pro- also highly conceivable synthetically.[54] The corresponding ceeded in high yields with high diastereomeric (de) selection ketone (72) initiated the first point of convergence, which was and enantiomeric excess (Scheme 9). To effect the second acyl facilitated by considering acyl anion equivalent chemistry (i.e. anion coupling, a second epoxide was required that contained 73),[30] which provided a left- (74) and right- (75) hand frag- syn-1,2-diol attributes (i.e. 79). Access to this unit (i.e. 93) arose ment. The left-hand fragment (74) could be further dissected from deployment of an enantiotopic group and diastereotopic using the same acyl anion approach to give epoxides 76 and 77, face-selective Sharpless epoxidation reported by Schreiber and whereas the right-hand fragment (75) was perfectly poised for a others[59] starting from divinyl carbinol (94). However, a ste- ring-closing metathesis-styled disconnection leading to 78, reochemical inversion of the first formed product (i.e. 95) was which could be accessed from a functionalised syn-diol of the required to obtain the desired stereochemistry, and this was type seen in 79. The convergent approach also provided several achieved using Mitsunobu conditions, which gave ester 96. The built-in fallback strategies associated with the left- (74) and ester was then cleaved, and TBS protection installed using right- (75) hand fragments. For example, the left could be t-butyldimethylsilyl chloride (TBSCl). Although the TMS- accessed via traditional asymmetric aldol methods (e.g. sequen- dithiane (87) was used initially, the TBS-dithiane anion (97) tially coupling 80 to aldehydes 81 and 82), whereas the right (i.e. treatment of TBS-dithiane (98) with nBuLi) was found to hand could materialise from the chiral pool (e.g. deoxy sugar 83) provide superior coupling in terms of delivering the advanced (Scheme 8). intermediate 99 (Scheme 9). In the forward direction, the left-hand fragment 74 was Completing the synthesis required considerable trouble- tackled first, as it was a chance to explore and select the shooting in terms of finding ring-closing metathesis (RCM) appropriate acyl anion chemistry for the final coupling of both conditions, and a workable sequence of deprotection and cycli- halves. In this regard, Tietze[55]–Smith[56] linchpin methodol- sation through to the target.[53,60,61] The final sequence pro- ogy was chosen, which is founded on the Brook rearrangement ceeded as follows. Addition of the acryloyl group to 99 of silylated dithiane anions that facilitate an anion relay reaction proceeded smoothly to afford 100, but the ensuing RCM reac- involving non-symmetrical epoxides. On that basis, epoxide 84 tion required both harsh conditions and a catalyst that could was first prepared. This was achieved in two steps (via 85) using maintain sustained activity. In this regard, the second- the method reported by Lepoittevin,[57] which provided the generation Hoveyda–Grubbs (101) catalyst[62] delivered in desired material in high enantiomeric excess (ee) and in large spades, in what is believed to be overcoming severe steric

H H RCM O + O H HO HO HO O OH OH O O O H O O 12 12 72 EBC-23 (71) Acyl anion equivalent

PO O O PO PO PO PO O FG FG ++XY FG FG OO 12 12 73 OO 82 74 75 78 81 R R 80 Left hand Right hand fragment fragment O O O FG AcO PO P = Protecting group 12 77 AcO FG = Functional group 76 XY FG 73 OAc OH 83 79

Scheme 8. EBC-23 (71) retrosynthetic analysis. EBC ¼ EcoBiotics Compound. 298 C. M. Williams and M. A. Dallaston

OH NaOH CH3(CH2)11MgBr O CI CI THF / MeOH 12 Cul / THF 86 91 % 85 (75 %)

nBuLi nBuLi TMSCI SS 90 % SS 87 TMS 88 O 86 O CI TMSO Hg(CIO4)2 OH O + SS SSO O 12 THF / Et2O CaCO3 84 TMS 12 90 12 61 % THF / H2O 91 (79 %) 89

1) Et2BOMe 2) Me2C(OMe)2 NaBH4 TsOH· pyr O DIAD/PPh O 1) K CO / MeOH OH 3 OCOAr 2 3 THF / MeOH CH2CI2 95pNO2PhCO2H 96 2) TBSCI / Imid 65 % >95 % O TiOiPr L-DIPT 4 93 OO tBuOOH 70–80 % TBS OTBS O O O O OH OH S S 12 92 (80 %) 12 SS 94 de >95:5, ee 92 % 99 (78 %) OTBS 97 TBS

nBuLi TBSCI nBuLi SS SS 88 98 TBS

Scheme 9. The double Tietze–Smith anion relay route to access advanced intermediate 99 (L-DIPT ¼ (þ)-Diisopropyl L-tartrate). encumbrance and likely counterproductive co-ordination to the natural product isolation field.[67] To highlight the point of dithiane sulfur atoms. The desired product (102) was then mechanistic insight, Woodward chose the santonin (104) and stripped of silyl protection using hydrogen fluoride (HF) to give santonic acid (105) story as an exemplar case.[67] For 100 years, the polyol (103), setting the stage for final spiroketalisation. many organic chemists had been trying to solve the structure of Cerium ammonium nitrate (CAN) was found to be critical for santonin (104),[68] which had been isolated from wormseed this final step in that it not only facilitated the unmasking of the (Artemisia santonica) in 1830.[69] Although the flat structure of keto functionality, but also metal-templated the cyclisation, as santonin (104) was solved at the end of this 100-year period, one observed previously by Evans.[63] This last manoeuvre deliv- of the unexplained products of degradation was santonic acid ered EBC-23 (71) concisely (i.e. two degrees of convergence (105).[67] Interestingly, this Holy Grail of the time was obtained with nine total linear steps in 8 % overall yield), which enabled by vigorous boiling of 104 with concentrated alkali.[67] Wood- confirmation of structure and further biological studies ward set about demonstrating that by evaluating the various (Scheme 10). base-induced degradation products and intermediates from a mechanistic view point, a proposed structure could be rationally Chemical Degradation determined, and then verified experimentally.[65,67] Woodward Chemical degradation covers an aspect of retrosynthetic focussed on the functionality that would be likely susceptible to thinking that was only really made possible owing to the treatment with hydroxide, which first involved ring-opening then-emerging field of natural product isolation and eluci- of the lactone in santonin (104) to give alcohol 106. Enolisation dation (i.e. before modern methods involving NMR of the conjugated g-enone in 106, via 107, revealed the possi- spectroscopy).[64] The main approach in the early era was to bility of a ketone (i.e. 108) in the C-ring. The C-ring ketone perform a multitude of chemical degradation studies on the could then undergo enolisation and reach over and attack the unknown isolate looking for functional group and skeletal clues enone in the A-ring (i.e. 109). It was this critical thinking that (e.g. via melting point, boiling point, density, refractive indices, led to the notion that santonic acid (105) was a cage bicyclic optical rotation) to assist in solving the structure.[64] Elucidation system (i.e. quite the novelty system in the day) (Scheme 11). methodology of this type took a great leap forward, however, Note it took many more decades for the absolute stereochem- when in 1948 Woodward combined the evolving mechanistic istry of both (–)-a-santonin (110)[70] and santonic acid (111)[71] understanding of chemical reactions with natural product elu- to be solved. cidation (i.e. understanding degradation pathways from a Although santonic acid (105) was not a natural product in its mechanistic standpoint).[65] By this time, of course, chemical own right, this endeavour married the emerging field of reaction reactions were becoming better understood through the adoption mechanism understanding with the field of natural product of reaction arrows; however, although having been introduced chemistry. The outcome of this merger produced several key 25 years prior,[66] these concepts were slow to be adopted in the discipline shifts: (1) it added a new dimension to the elucidation The Future of Retrosynthesis 299

CI O TBS TBS O O OOH O O O O O SS S S NEt3 / CH2CI2 12 82 % 12 99 OTBS 100 OTBS

MesNN Mes CI 101 Ru µW / Toluene CI 150°C / 4 h O 65 % iPr

O O TBS HO HO HO O HF O OO O SS SS CH CN / CH CI 12 3 2 2 12 103 OH 102 OTBS CH CN / H O 3 2 CAN 54 % 2 steps

H O OH O OH O O H 12 (+)-EBC-23 (71)

Scheme 10. Completion of the EBC-23 (71) synthesis.

O O OH O

H2O HO¯ OH H2O OH O O H H

CO2¯COCO2¯ 2¯ O 108 104 106 107 HO¯ O

O O O H O CO2H O O O H O¯ (–)--Santonin Santonic Acid CO2¯ (111) 105 CO2¯ (110) 109

Scheme 11. Woodward’s postulate for the base-induced conversion of santonin (104) to santonic acid (105). Flat structures presented to reflect the knowledge of the day.

of natural products (and degradation products); (2) it elevated in northern Australia, Indonesia, and Papua New Guinea, and the notion that natural products could be transformed into have been used extensively by native tribes owing to their related natural products and/or systems; (3) it assisted in laying hallucinogenic and medicinal properties.[72–75] the foundation for partial and relay syntheses (see below); and Himgaline (112) is an interesting case in point.[72,76,77] It (4) it demonstrated that new chemical transformations could be was solved using classical chemical degradation reinforced by incorporated into retrosynthetic thinking. semi-synthesis, with the final stages of the structure being A further example to highlight these points are the Galbuli- proved using oxidative elimination of 112 using 70 % nitric mima alkaloids, which also integrated biosynthetic understand- acid to give GB13 (113). The overall conversion was repre- ing. These unique alkaloids, which now number ,35 members sented in several simplified schemes showing the starting in four separate classes, have so far been isolated from the bark alcohol moiety (114) giving the eliminated enone function of the tropical rainforest trees Galbulimima belgraveana and (i.e. 115) and the fragmented dialkylamine (116).[78,79] Further- Galbulimima baccata.[72–75] These evergreen species are found more, it was reported that catalytic hydrogenation of GB13 300 C. M. Williams and M. A. Dallaston

(113) in the presence of acid reconstituted himgaline (112), converted into oximes (i.e. 132), and then reduced and cyclised both installing the last nitrogen-containing ring (via an as an annulation tactic to install the last ring of the GB13 aza-Michael addition) giving oxohimgaline (not shown), and skeleton (i.e. 133). Skeletal manipulations and protecting group hydrogenating the resulting ketone to an alcohol (i.e. dancing afforded ketone 134, which after Saegusa oxidation himgaline).[78,79] In light of this information, it was clear that created the necessary enone moiety seen in 135. Double depro- any total synthesis of himgaline would be best achieved via tection of the piperidine ring nitrogen and the bridgehead GB13 (Scheme 12). hydroxy group provided GB13 (113) in a total of 29 steps Mander and McLachlan were the first to report a synthesis (Scheme 13). of ()-GB13, via a sequence that involved multiple dearo- At the completion of the synthesis, Mander and McLachlan matisation steps, ring contraction/cleavage/annulation, and a stated that the synthesis of GB13 laid a foundation for the Diels–Alder cycloaddition.[80] The details of this synthesis, preparation of himgaline (112).[80] They further commented which showcases the utility of benzenoid synthons,[81] started that intramolecular Michael addition of the nitrogen to the enone with a Birch reduction[82] alkylation protocol.[83] The dianion of in GB13 followed by reduction of the ensuing saturated cyclic 2,5-dimethoxybenzoic acid (117) was generated via lithium– ketone should lead to himgaline, according to the degradation ammonia reduction, and then quenched with 3-methoxybenzyl studies reported by Mander and Prager et al. in 1967.[78,79] bromide (118) to give the dearomatised bis-enol ether 119. Subsequent work by Chackalamannil,[84] and Evans,[85] Double protonolysis and cyclisation of 119 with sulfuric acid achieved the total synthesis of himgaline (112) via different gave a benzobicyclo[3.3.1]nonane (120), which underwent routes, although the final stages of both syntheses were based on decarboxylation followed by methoxymethyl (MOM) protec- a retrosynthetic analysis involving GB13 (113) as the penulti- tion of the bridgehead hydroxyl (i.e. 121). Installation of the mate step before Mander’s predicted Michael addition reduction adjacent diazo function with p-nitrobenzenesulfonyl azide gave sequence. For other Galbulimima alkaloid syntheses (e.g. ketone 122, setting the stage for a Wolff rearrangement- (–)-GB17), see Thomson et al.[86] mediated ring contraction. Irradiation of 122 in the presence Additional examples that are recommended to the interested [87] of hexamethyldisilazane resulted in the intermediate ketene reader include the total syntheses of bleomycin A2 (136), being intercepted, and on protonolysis afforded the amide based on degradation-inspired retrosynthesis, and xylopinine (123). Even though a mixture of diastereomers eventuated, this (137),[88] based on a mass spectrometry degradative was inconsequential given that the next step aimed to install a fragmentation-inspired retrosynthesis. However, it should be double bond for the envisaged Diels–Alder reaction. This was noted that there are instances where complex unknown sub- achieved first by conversion of the amide function into a cyano stances have resisted elucidation by natural product degrada- group, and then via deprotonation and trapping with diphenyl- tion, e.g. phorbol (138)(Fig. 1),[89] and in turn attracted selenide, providing the option to install the bridge double bond considerable synthetic attention.[90] (i.e. 124)bysyn-elimination of the phenylselenyloxide. A normal demand Diels–Alder reaction involving the nitrile (124) and silyl enol ether 125 proceeded smoothly albeit slowly Partial Synthesis under Lewis acid catalysis (126) to produce the desired adduct Partial synthesis (or semi-synthesis) relates to natural product (i.e. 127). Interestingly, this reaction was performed neat but synthesis using a mixture of biosynthetic, degradation, and using a large excess of the silyl enol ether, i.e. the diene acted as FGI-inspired retrosynthetic concepts. This often means that both reagent and solvent. TBS deprotection, reduction, and the skeleton is already (or mostly) intact and that FGI protection afforded 128, which was subjected to a second chemistry is then deployed to achieve the synthesis of the dissolving metal reduction (keto and cyano) to unmask the target. In some cases, the starting point (i.e. starting material) enone (129) via liberation of the intermediate enol ether. can be non-natural, but ultimately it is also derived from a Stepwise epoxidation of 129 provided epoxyketone 130, which nature source (e.g. via degradation). A wonderful example underwent Eschenmoser fragmentation in high yield to give where the wide utility of semi-syntheses can contribute to 131. Both the acetylenic unit and the keto function were then an area of science is that of the gibberellin plant hormones.[91]

OH OH

OH HO H H N N H H H N H H OH O Himgaline (112) GB13 (113)

H HNO3 C C C N CCC H N H H HO 2 O 116 114 115

Scheme 12. Himgaline (112) structure confirmed by degradation and reconstitution from GB13 (113). The Future of Retrosynthesis 301

O OMe OMe CO2H MeO HO C MeO 1) AcOH / H2O / O i) Li / NH MeO 2 H SO (60 %) O HO 3 2 4 2) MOMCI / DMAP MeO Acetone, 0°C DIEA / DCM Br HO MOMO ii) 80 % 86 % over 2 steps 121 117 OMe 118 119 OMe 120 75 % 2) NEt3 / ACN 1) NaH EtOCHO O N SO N 2 2 3 THF

MeO 1) CI CCOCI / NEt MeO NH MeO 3 3 2 O Yb(thd)3 N DCM (98 %) hv / THF / HNTMS2 O 110°C 2) LDA / tBuOK then 2 M HCI / 0°C MOMO MOMO MOMO N2 3 days 124 Ph2Se2 / THF 123 68 % over 3 steps 122 87 % then H2O2 / 0°C (6:1 Exo : Endo) OTBS 74 % 125

MeO CN 1) TBAF / THF (74 %) MeO CN O H H 2) LiAIH4 / THF (94 %) H i) Li / NH3 / THF H H H 3) MOMCI / DMAP ii) HCI / MeOH MOMO H MOMO H MOMO H DIEA / DCM (96 %) THF (55 %) 127 OTBS 128 OMOM 129 OMOM

1) LiAIH4 / THF 3) DMP 2) mCPBA / DCM NaHCO3 77 % DCM

OH O O O N H O2NSO2NHNH2 H H H2NOH·HCI H H H Pyr / 100°C Pyr / EtOH / THF H H H MOMO 76 % MOMO N MOMO H H H HO 131 OMOM OMOM 132 OMOM 130

1) NaBH4 / ZrCI4 / THF 1) LDA / TMSCI / THF F3COC 2) Zn / AcOH, Et2O H O 2) Pd(OAc) / DMSO N 3) TFAA / NEt3 / DCM 2 H O ACN H Yb OO 82 % over 2 steps H O 1) 1 M HCI O F3COC F COC MOMO H Acetone / 3 H 135 N N O Yb(thd) (126) H 2) DMP / DCM H 3 1) K CO / H O 2) 1 M HCI H H 2 3 2 H 3) MOMCI / DMAP H THF / MeOH / Acetone / MOMO H DIEA / DCM MOMO H 133 OMOM 96 % 134 O GB13 (113) 32 % over 4 steps 57 % over 3 steps 37 % over 2 steps

Scheme 13. Mander’s total synthesis of ()-GB13.

NH H 2 H + H2N N NH2 Bleomycin A2 (136) O SMe2 O O N – NN H O S N X H H H O O N N S H2N O N H H HN O N O MeO H H H HO OH O N N HO MeO HO H O N O H H OH H Xylopinine HO O OH OH (137) OMe HO O OMe OOHOH O NH 2 Phorbol (138)

Fig. 1. Bleomycin A2 (136), xylopinine (137), and phorbol (138). Atoms highlighted in green represent sites for degradation and retrosynthetic disconnections.

A specific example is the conversion of gibberellic acid (139)to bamboo shoots (i.e. 14 mg from 44 tons), and was required for [92] [92] gibberellin A19 (GA19, 140). At the time, GA19 had only ever biosynthetic studies. First, gibberellic acid (139) was ester- been isolated once in workable analytical quantities from ified with diazomethane followed by protection of the 302 C. M. Williams and M. A. Dallaston bridgehead hydroxyl (via MOM) to give 141. Dissolving metal enabled by a relay sequence obtained from degradation of 152 reduction cleaved the lactone bridge, removed the A ring oxy- (via 157). The ability to inject a vital advanced intermediate into genation, and isomerised the double bond to afford 142, which the total synthesis campaign was critical to the completion of the was converted into the diazoketone (143). Cyclopropanation planned route, which involved re-establishment of the furanone mediated by copper was the key synthetic manoeuvre to intro- ring via 158 and 159 (Scheme 15). duce the desired carbon bridge, and as expected, the cyclopro- pylketone (144) was obtained in high yield. A second lithium in Chemoenzymatic Synthesis ammonia reduction cleaved the cyclopropane bond that made up Building on statements made above concerning biomimetic a six-membered ring, in turn leaving a five-membered ring syntheses, it was noted that enzymes are an important set of ketone (145). Oxidative cleavage of the resulting five- ‘reagents’ that can be integrated solely or partly into retro- membered ring was challenging, but it eventually yielded to a synthetic route design. A prime example is enzyme-mediated two-step process first involving treatment with potassium hydroxylation of hydrocarbon scaffolds using cytochromes hydride (KH) and quenching with methyl iodide to afford an P450 (P450s).[100] To highlight this point, Shen and Renata took enol ether (146), followed by exposure to ozone. Reductive commonly available diterpenes and converted them into high- workup with dimethylsulfide (DMS) provided the desired ter- value and rare natural products to demonstrate the power of tiary aldehyde and C4 carboxylic acid functional groups (i.e. combining chemical and enzymatic oxidation methods via a 147). Sequential basic and acidic hydrolytic conditions revealed chemoenzymatic platform.[101] Among the numerous examples, the target (Scheme 14). steviol (160) was first oxidised with PtmO6, which was better Pharmaceutically relevant examples include the semi- overproduced on expression in Escherichia coli, followed by a synthesis of both TaxolÒ (148)[93] and artemether (149) [94] traditional pyridinium dichromate (PDC) oxidation to give the (Fig. 2). C7 ketone (161). Not only was the yield high over two steps, but the selectivity was remarkable. Relay Synthesis Note: regioselective C7 oxidation of terpenoid substrates is Relay synthesis is a term used for the conversion of one natural challenging, as exemplified by attempts to construct the ABC product into another so as to intercept and supply an advanced ring system (i.e. 162) of gedunin (163) via the devised intermediate for a total synthesis campaign. For retrosynthetic thinking, a route that employs a relay overcomes access to advanced intermediates in meagre quantities via chemical syn- O H O thesis by using partial synthesis. A very early example was O OH O provided by Robinson for the synthesis of cholesterol (150),[95] O O O NH O HH with more recent examples reported by Shi, Tan, and Gin for the O O O synthesis of the C -bisnorditerpene alkaloid neofinaconitine O O 18 OH HO Ac OMe (151),[96–98] and Deslongchamps et al. for ouabagenin (152).[99] O Artemether The latter is a particularly instructive example whereby Taxol® (148) (149) advanced intermediate 155 was achieved via the planned Diels– Alder based retrosynthetic route (i.e. reaction of 153 with 154 to Fig. 2. TaxolÒ (148) and artemether (149). Atoms highlighted in green give 155), but additional material (i.e. 156) was produced, reflect sites of retrosynthetic disconnections.

O H O HH OH 1) CH2N2 / Et2O OMOMLi / NH3 OMOM CO CO 2) MOMCI / DMAP tBuOH HO DIEA / DCM MOMO 93 % H CO H H CO Me H CO Me 142 2 2 HO2C 2 139 141 (CICO)2 / Pyr then CH2N2 DMF 83 %

H H H OMOM Li / NH3 OMOM Cu OMOM CO tBuOH THF / C6H12 H 93 % H H CO2Me CO Me CO Me O 2 N2HCOC 2 145 144 (87 %) 143 KH then DMF Mel

OHC H H OHC H 1) KOH / OH OMOM O3 / pyr OMOM MeOH / H2O MeO 2) 3 M HCI CHCI3 / MeOH H CO H H then DMS H CHCI / MeOH HO2C 2 CO2Me HO C CO2Me 3 2 GA19 (140) 146 (65 %) 147

Scheme 14. Partial synthesis of gibberellin A19 (GA19, 140). The Future of Retrosynthesis 303

OTBDPS OTBDPS OHC OAc O O AcO O Cs CO + 2 3 O O DCM / 0°C H PhMe Si 155 2 154 OH 85 % PhMe Si O 153 2 H O O O O

O OH O AcO O AcO OTBDPS HO OH OAc OAc 4-MeOC H CO HO 6 4 2 4-MeOC6H4CO2 H H H H OH H OH H OH 156 HO AcO AcO OH OH 157 OH Ouabagenin (152) nBu nBu O Sn O O O AcO OAcAcO OAc Na2CO3 4-MeOC H CO 4-MeOC H CO 6 4 2 H 6 4 2 H MeOH 85 % H OH H OH 159 158 AcO AcO OH OH

H OMe OMe N OMe H H OH HH H O HO Neofinaconitine Cholesterol (150) NH2 O (151)

Scheme 15. Relay synthesis of ouabagenin (152) by Deslongchamps et al. retrosynthetic plan involving advanced intermediate 164. After the advanced intermediate of choice becomes the synthetic considerable investigation, the synthesis of 162 was achieved in target. These indirect targets are usually embedded in the final 17 steps starting from 165 (Scheme 16).[102] stages of a prior reported synthetic route. Continuation of the chemoenzymatic synthesis involved Formal syntheses are numerous and widespread. An instruc- another enzymatic oxidation, but this time using PtmO5, which tive example is that of the marine natural product 7,20-diiso- is a class I P450 that requires a separate reductase partner cyanoadociane (170), which became a popular target because of (e.g. RhFRed) to support function and turnover. The adoption considerable antimalarial activity shown by this and related of this enzyme installed a b-OH at C11 with absolute stereo- systems.[104] Corey was the first to complete a total synthesis of specificity to give 166. Acylation of 166 and deployment of the 170 in 26 steps, via the dione advanced intermediate 171, Appel reaction gave amide 167, which underwent double starting from glutaric anhydride (172).[105] Mander was the first hydride reduction to give 168. Subjecting 168 to selenium to establish the stereochemical relationships at C7 and C20, by dioxide gave the bridge allylic alcohol (i.e. C15) that was completing the first formal synthesis of 170 starting from the selectively oxidised with IBX[103] to afford rosthornin C (169) anisole derivative 173.[106] The intersecting target in this formal (Scheme 16). synthesis was the diamine (174), which Simpson and Garson had previously reported could be converted in two steps to 170 (Scheme 17).[107] However, numerous groups focussed on Formal Synthesis targeting Corey’s dione (i.e. 171) as the point of establishing A formal synthesis is achieved when a practitioner performs a their formal syntheses. The first was that by Miyaoka,[108] which synthesis of a known target molecule using a different approach took 29 steps from (S)-4-methyltetrahydro-2H-pyran-2-one to that reported, but instead of completing the entire synthesis, (175), followed by Vanderwal in 21 steps from perillaldehyde the sequence stops at an advanced intermediate intersecting the (176),[109] and then Thomson with a shattering 12-step synthesis previously disclosed synthetic route. For the purposes of retro- of the dione (171) originating from desmethylcarvone synthetic thinking, it is not the target itself that warrants a full (177).[110] Vanderwal ended up finishing this entertaining fight disconnection evaluation, and associated forward planning, but with a knockout blow constituting a 10-step total synthesis 304 C. M. Williams and M. A. Dallaston

(a) OH HO OH OH 1) PtmO6 / αKG 11 O 2 / PtmO5- 2+ Fe / O2 RhFRed H H H 7 2) PDC O NADP / Opt13 O 85 % over 2 steps H 65 % H H HO2C HO2C HO C Steviol 161 166 2 (160) Enzymatic Enzymatic Hydroxylation 1 Hydroxylation 2 1) Ac2O

2) PPh3 / I2 Imid

AcO OH AcO OH AcO OH

1) SeO2 / TBHP NaBH4 15 H H H 2) IBX 7 59 % O OH over 2 steps O H OH 66 % H H over 2 steps 19 HO Rosthornin C HO 168 N O 167 (169) N

(b) O O

O OH 17 steps O H O O 7 O H O H O OH O H O OAc O OP H H 162 165 Gedunin (163) 164

Scheme 16. (a) Chemoenzymatic synthesis of rosthornin C (169) from steviol (160); and (b) retrosynthesis of gedunin (163) with construction of the advanced intermediate (162). aKG ¼ a-ketoglutaric acid; NADP ¼ nicotinamide adenine dinucleotide phosphate; Opt13 ¼ thermostable phosphite dehydrogenase variant; Imid ¼ imidazole; TBHP ¼ tert-butyl hydrogenperoxide.

Garson 2 steps Corey 3 steps

O H NH2 H NC H 171 7 O H H H H H H Corey O O H2N H H CN H H 20 H H 23 steps 172 O H H H O 174 7,20-Diisocyanoadociane Miyaoka (170) O 175 29 steps Mander 42 steps Vanderwal Thomson 2nd Gen CHO 12 steps Vanderwal 10 steps 1st Gen 21 steps OMe H H 176 CO2Me O O 173 178 177

Scheme 17. Formal and total syntheses of 7,20-diisocyanoadociane (170).

(not formal) of 170 starting from dehydrocryptone (178) with certain reaction types (e.g. Claisen rearrangement);[112] (Scheme 17).[111] and (3) trust (or even distrust) of certain synthetic methodolo- gies (Suzuki cross-couplings).[9] Overall, a key piece of meth- Methodology-Driven Synthesis odology is identified as the focal point of the synthetic route Many target-orientated syntheses are retrosynthetically design, and this in turn impacts the retrosynthetic thinking designed around a certain type of synthetic methodology. This process, as the prime task then becomes strategic implementa- choice can be inspired by a multitude of reasons, some of which tion (and designing accommodation) to reach the target. include: (1) testing newly discovered methodology designed in- A synthetic campaign that encapsulated aspects of house (e.g. C–H activation);[10] (2) long-standing experience methodology-driven synthesis concepts was that of vibsanin The Future of Retrosynthesis 305

E(179)[113] and 5-epi-vibsanin (180),[114] which are two rarely mixture of syn-(196)andanti-isomers (197). The syn-isomer occurring complex cage-containing diterpenes isolated from (196) was taken forward through a sequence of deprotection, Viburnum awabuki in Japan.[115] Initially, the synthetic methods alcohol oxidation, and Wacker oxidation to unmask the penulti- of choice combined an acid-catalysed cyclisation of 181, fol- mate tricarbonyl 198. The last, and perhaps hopeful, step was lowed by ring expansion of 182, to give an advanced intermedi- heavily reliant on in-house-optimised Anders–Wittig chemistry ate containing features seen in 183.[27] However, although this (Anders–Gaßner reaction),[119] which delivered the target (i.e. work provided substantial experience for the neovibsanins (see 180 as the corresponding enantiomer 200), albeit in a low yield Schemes 3 and 4), it was not possible to convert the advanced of 26 % using the Wittig reagent 199 (Scheme 19).[114] intermediate into either of the targets.[116] Therefore, a more Although the Claisen rearrangement gave a mixture of syn- modern methodology was employed. However, it is always a and anti-isomers, it was possible to convert the anti-isomer into concern when adopting new methods as many often fail when vibsanin E (179)(Fig. 3), although this was only undertaken in applied to complex systems. In this case, however, the Davies the racemic series.[113] However, although a well-devised syn- group had shown that donor–acceptor diazo functions (e.g. 184) thetic route was in hand for these systems, it was not guaranteed undergo rhodium-catalysed [4þ3]-cycloaddition with dienes that other members of the vibsane family could be conquered, (e.g. 185) (a cyclopropanation Cope rearrangement sequence) and a spectacular failure was endured with attempts to synthe- to give seven-membered rings (e.g. 186). In addition, they had sise 3-hydroxyvibsanin E (201). The synthetic route above that applied this protocol to the total synthesis of ()-5,10-bis-epi- was arduously devised for vibsanin E and 5-epi-vibsanin E came vibsanin E (187), which was a non-natural product.[117] Clearly, completely unstuck in the penultimate step in the campaign to combining the [4þ3] and [4þ2]-cycloaddition methodology 3-hydroxyvibsanin E, when t-butyldimethylsilyl deprotection was very powerful, but the end game was compromised. This could not be achieved, giving 3-OTBS-vibsanin E (202) opened an opportunity for the Williams and Davies groups to (Fig. 3).[120] collaborate and pool experiences to complete the synthesis, by adopting the Davies protocol to access an advanced intermediate Collective Total Synthesis of type 183, and then deploying Williams end-game know-how The term collective total synthesis is a fairly new concept that (Scheme 18).[118] was first put forward by MacMillan in 2011, when his group Reducing this approach to practice, the initial seven- achieved the synthesis of six alkaloids from the same advanced membered ring (i.e. 188) was obtained in 65 % yield using the intermediate.[121] MacMillan argued that although syntheses chiral catalyst known as PTAD (189), which required deoxygen- targeting an advanced core structure applicable to the synthesis ation using in part Comin’s reagent (190) to afford a suitably of closely related natural products within a family had been substituted ring (i.e. 191) poised for the [4þ2] hetero cycloaddi- previously reported, it was much less common to find examples tion. The Diels–Alder reaction proceeded as planned and the where the preparation of an advanced intermediate with resulting tricyclic ring system (192) was converted to the key appropriate functionality was amenable to the construction of advanced intermediate 193 in three steps. Conjugate addition structurally diverse natural products in different families, the using the lithiated MOM derivative took place both in high yield latter constituting the definition of the term collective total and stereoselectively. Introducing TMSCl was critical to facili- synthesis, which has analogies with a biosynthetic pathway tating the reaction and providing the desired material (194). providing collections of metabolites.[121] Since that time, further Transmetallation and quenching with allyl bromide gave the allyl examples using different natural product classes have been ether (195), and subsequent Claisen rearrangement afforded a added,[122] but the term is somewhat contentious, because there

O OH O O O

20 181 182

-Alkylation Ring Expand O expansion Hetero O O O O O [4+2] O 5 10 R 1,4-Addition 183 Vibsanin E (5 , 10 , 179) [4+2] 5-epi-Vibsanin E (5, 10, 180) 5,10-bis-epi-Vibsanin E (5, 10, 187) O Formal [4+3] CO Me O 2 + N2 OTBS 184 185 186 (R = H)

Scheme 18. Retrosynthetic analysis of vibsanin E (179) and certain isomers. 306 C. M. Williams and M. A. Dallaston

R CO Me CO Me Rh2( -PTAD)4 2 2 CO2Me 189 1) TBAF / ACN + N2 OTBS OTBS Toluene 2) 190 / NaH / THF –15°C to 185 184 65 % 188 3) Bu3SnH / Pd(PPh3)4 191 (70 %) (90 % ee) LiCl / THF

1) DIBALH / THF 3) BF3·Et2O 2) DMSO / (COCl)2 DCM / –78°C DCM OTMS O

O MOMOCH2Li O O 1) NaCNBH3 / AcOH Cul / TMSCl 2) SeO2 / Dioxane MOMO TMEDA / THF 193 3) PCC / DCM 194 91 % 192 (77 %) 60 % Allyl bromide MeLi / THF 74 % HMPA O O O O O W, 185°C O 1) 10 M HCl, MeOH O Toluene 2) DMSO / (COCl) MOMO 2 MOMO DCM O 195 Syn 41 % (196) 3) PdCl / CuCl / O 198 (20 %) Anti 197 2 2 2 11 % ( ) DMF

Cl O O O O 199 N ORh O O N NTf2 O PPh3 O Rh Comin’s reagent O 190 NaNTMS2 / THF 4 ( ) O 26 % R epi Rh2( -PTAD)4 (–)-5- -Vibsanin E 200 (189) ( )

Scheme 19. Total synthesis of (–)-5-epi-Vibsanin E (200). Rh2(R-PTAD)4 ¼ tetrakis[(R)-(-)-(1-adamantyl)-(N-phthalimido)acetato]dirhodium(II); TBAF ¼ tetra-N-butylammonium fluoride; ACN ¼ acetonitrile; DIBALH ¼ diisobutylaluminum hydride; MOMOCH2Li ¼ ((methoxymethoxy)- methyl)lithium; TMEDA ¼ tetramethylethylenediamine; HMPA ¼ hexamethylphosphoramide; NaNTMS2 ¼ sodium bis(trimethylsilyl)amide; PCC ¼ pyridinium chlorochromate.

O O O O OH O O 5 5 3 O 10 O 10 O O Vibsanin E (179) 3-Hydroxyvibsanin E (201)

O O OTBS O 5 3 O 10 O 3-OTBS-vibsanin E (202)

Fig. 3. Vibsanin E (179), 3-hydroxyvibsanin E (201), and derivative 202. is an element of false pretences (i.e. the deck can be stacked by performed such that there is sufficient choice to undertake such making a judicious choice of targets to match the key advanced an endeavour. Overall, however, the concept has substantial intermediate). Alternatively, a collective total synthesis can merit in that it aims to make one advanced intermediate multi- only be achieved if enough natural product elucidation has been task for multiple target syntheses, which has several important The Future of Retrosynthesis 307 advantages. For example, providing a wide array of targets for asymmetric form using the following sequence, the only differ- biological testing saves time and effort designing individual ence being that in the racemic procedure, a Mukiyama aldol was synthetic routes, and is green chemistry-orientated (i.e. limiting deployed, whereas the enantioselective route employed a boron- waste streams). mediated aldol using (þ)-DIP-Cl (215). In brief, treating alde- The tetranortriterpenes (limonoids) are such a natural prod- hyde 218 with the boron enolate of 2-butanone, followed by uct class that fit the collective total synthesis criteria, in that a potassium hydride-induced cyclisation gave the desired cyclo- huge amount of structure elucidation has been undertaken on hexanone (219), which was easily converted into the required many different species within the meliaceae and rutaceae fami- enol ether (217) poised for condensation and rearrangement. lies.[123] A case of serendipitous collective total synthesis is that Heating a mixture of 216 and 217 in the presence of anhydrous of the bicyclononanolides (e.g. mexicanolide (203)), whereby tosic acid afforded, via the extended enol ether 220, the second the retrosynthetic analysis aimed to access azedaralide (204)as natural product, cipadonoid B (208), as the major product.[124] the key advanced intermediate to achieve a total synthesis of Epoxidation of cipadonoid B to introduce the additional ring mexicanolide (203), via a pre-established proposed biosynthetic oxygenation proceeded smoothly. Interestingly, it was not route (i.e. via 205). Entry to azedaralide (204) was envisioned to possible to reduce the epoxide to access the proposed inter- be directly achievable from pyroangolensolide (206), a known mediate 205 (Scheme 20) using traditional reagents. However, degradation product that had been synthesised several times treatment of the epoxide (221) with aluminium amalgam not previously, and could be obtained starting from 2,6-dimethyl- only regioselectively reduced the epoxide ring, but the resulting cyclohexenone (207)(Scheme 20). In this case not only were enolate subsequently underwent a 1,6-conjugate addition to two limonoid natural products synthesised (i.e. 203 and 204), afford the third natural product, proceranolide (209). This but an additional three were also achieved (i.e. 208–210). fortuitous result opened the opportunity to synthesise the fourth In the forward direction, the asymmetric synthesis of natural product, khayasin (210), using a peptide coupling (þ)-azedaralide (216) followed the racemic synthetic route protocol to install the hindered isopropyl ester. In addition, the (i.e. 204).[38] First, cyclohexenone 211 was subjected to original target and the fifth natural product, mexicanolide (203), in-house optimised Morita–Baylis–Hillman conditions,[29] could also be synthesised from proceranolide (209) using the which afforded the extended ketone (212) after TBS pro- Jones oxidation protocol (Scheme 21).[125] tection. Methylation (i.e. 213), followed by a ()-diisopinocampheylchloroborane- (see 215 for (þ)-DIP-Cl Chemical Database-Aided Synthesis isomer; DIP ¼ diisopinocampheylborane) mediated asymmetric Several searchable chemical databases are available to the diastereoselective aldol, provided the furylalcohol (214). The practitioner. Reaxys and SciFinder are the main staple in this alcohol function was then acetylated to facilitate cyclisation, regard. Key features include the capability to find specific followed by TBS deprotection, to give the first natural product compounds and substructures, general and structure specific as the advanced intermediate (i.e. (þ)-azedaralide (216)). A reactions, physical properties (experimental and calculated), rarely utilised ketal-Claisen rearrangement was used to attach and spectroscopic data (experimentally determined and the A and B ring components of the targets to azedaralide. For calculated). this retrosynthetic manoeuvre to even be considered, a suitably In the context of database searching as applied to retro- functionalised cyclohexenane ring, matching the features seen synthesis, D’Angelo and Smith put forward the notion that in 217, was required. This goal was achieved in both racemic and retrosynthesis should be used to initially find a product from

O Mexicanolide O O (203) 7 species O O O O MeO2C H O O O O 30 2 H 2 O 30 205 HO Azedaralide CO Me O 2 (204) 1 species O O

O O O O MeO2C H O O O H O O CO Me OR 2 O Proceranolide (209, R = H) Cipadonoid B 207 10 species Pyroangolensolide (208) (206) Khayasin (210, R = COiPr) 1 species 5 species

Scheme 20. Retrosynthesis of the mexicanolide group. Numbers of species that have been found to produce certain members of this group are highlighted in dark blue. 308 C. M. Williams and M. A. Dallaston

1) Formalin LDA / THF KHMDS / THF MeI OH O DMAP / SDS O O (−)-DIP-Cl 2) TBSCl / DCM 64 % 211 TBSO 212 TBSO 213 Furylaldehyde O NEt 44 % 3 214 74 % 2 steps TBSO

1) Ac2O / DMAP / pyr 3) TBAF 2) LDA / THF THF O O O OMe MeO C TsOH 180°C O 2 O O + O Xylenes O O H O MeO2C O 217 HO CO Me 220 2 (+)-216 (21 %) Cipadonoid B MeOTf / DCM 90°C (208, Major) O O 1) (+)-DIP-Cl / DIPEA MeO C 2-Butanone 2 H 2) KH / Toluene 219 H2O2 / K2CO3 218 75 % 33 % 2 steps MeO2C H2O / MeOH O O O

O O O Isobutyric O Al/Hg MeO2C H acid MeO2C H O O O O H EtOH / THF O EDCI / DMAP O H2O / NaHCO3 DCM ))) CO2Me 221 OH OCOiPr Proceranolide Khayasin (209, 30 %) (210, 71 %) O Mexicanolide (203, 68 %) K2Cr2O7 H2SO4 MeO C H O Cl 2 Acetone O B O

(+)-215 O

Scheme 21. The collective total synthesis of the five limonoid natural products 203, 216, and 208–210. ))) ¼ sonication. undertaking a disconnection, which then enables a computer chemical structures and atom connectivity in a way that was search to generate the remainder of the retrosynthesis. They machine-readable. The most commonly used method is molec- termed this process as ‘a hybrid retrosynthesis approach’, ular graph theory, in which the molecule is represented by nodes because the concept utilises the basics of retrosynthetic analysis and edges. Feature matrices are then used to encode various to identify a working intermediate that is subsequently searched properties of the atoms (i.e. nodes) and relationships or bonds for in a database using a computer.[126] between them (i.e. edges), meaning that a 2D graph can encode 3D information such as stereochemistry.[130] Graph theory has Computer-Aided Synthesis and Artificial Intelligence been particularly useful in describing, and exploring, the small- The use of computers in chemistry has been around nearly as molecule chemical universe.[131] long as personal computers themselves,[127] and their possible At its heart, retrosynthetic analysis can be seen as a pattern application to planning synthetic routes has been realised for recognition exercise, an activity that computers are exceptionally nearly as long.[128] There have been several comprehensive good at. A set (albeit an ever-expanding set) of fairly simple rules reviews on the topic,[13c,129] and so here we will only cover the can be used to solve even incredibly complex problems. In this topic briefly and highlight a few key programs and ideas. A regard, the parallels between retrosynthetic analysis and chess selection of past and present programs is listed in Table 1, which have been discussed before.[129a] In 1997, a computer called by no means covers every program in the literature. Deep Blue famously beat the world chess champion in a six- One of the first challenges to overcome in the realm of game match after making a move that has variously been computers in chemistry was to develop a method to code attributed to a bug in the software and to superior The Future of Retrosynthesis 309

Table 1. A selection of past and present retrosynthetic, forward prediction, and starting material selection computer programs Information was gathered from the cited literature

Name Type YearA Notes External database Readily available

OCSS/LHASA[12a] Retrosynthetic 1969 Human-coded rules None No SYNCHEM[13a] Retrosynthetic 1977 Knowledge based, heuristic search, Aldrich Chemical Co. and No non-interactive common compounds SECS[13b] Forward 1978 Logic-centred, knowledge based, None No prediction heuristic search EROS[138] Forward 1978 Formal reaction generators that regard None No prediction reactions as bond- and electron- shifting processes CAMEO[139] Forward 1980 Human-programmed mechanistic rules None No prediction CHIRON[140] Starting 1984 Analyses input for stereochemical Manually compiled compound No material information and searches a database catalogue with well-defined of starting materials for matching stereochemical features compounds SST[141] Starting 1984 Hierarchical search within a starting A portion of the Aldrich No material material library catalogue SYNGEN[142] Retrosynthetic 1989 Programmed rules to mechanistically REACCS, SYNLIB No test all possible reactions RETROSYN[143] Retrosynthetic 1990 Reaction classes from a database to A subset of the ORGSYN No suggest reactions Database of Molecular Design WODCA[144] Retrosynthetic 1990 Logic-oriented heuristics, interfaced Built in starting material No with EROS for forward prediction database Beppe[145] Forward 1992 Classification by similarity, based on None No prediction modelling SOPHIA[146] Forward 1995 Heuristic approach SYNLIB plus an original data- No prediction base constructed from text- book reactions SESAM[147] Starting 1998 Input targets and starting materials are None No material described in connectivity tables, considers only the carbon skeleton KOSP[148] Retrosynthetic 1999 Reaction knowledge base derived auto- AIPHOS, Available Chemicals No matically from reaction database Directory (ACD) ROBIA[149] Forward 2005 Programmed rules and molecular None No prediction modelling ICSynth[150] Retrosynthetic 2005–2013, Java applet; Machine learning to generate sets of SPRESI, option to incorporate Yes – commercial 2014 onwards, HTML5 chemical rules from reaction others databases ReactionPredictor[151] Forward 2012 Machine learning approach, Hand-curated from literature Yes prediction mechanistic and graduate level textbooks Chematica/ Retro and 2013 (as a poster presen- Algorithms that draw from a database Network of Organic Chemistry, Yes – commercial Synthia[152] forward tation at the 245th ACS of hand-coded rules MilliporeSigma reagent national meeting)[153] database Reaction prediction Retrosynthetic 2017 Machine learning, Seq2Seq USPTO Yes seq_2_seq[154] IBM RXN[155] Forward predic- 2018 Natural language approach, Seq2Seq, USPTO[156] Yes tion and trained on automatically extracted one-step reaction data retrosynthesis LillyMol[157] Retrosynthetic 2018 Machine learning, trains reaction USPTO, option to use in-house Yes transformation rules, requires databases atom mapping Spaya AI[158] Retrosynthetic 2020 Deep machine learning Pistachio database Mcule, Yes – exited beta EMolecules, Chemspace in 2021 libraries AutoSynRoute[12e] Retrosynthetic 2020 Transformer-based Seq2Seq model and USPTO (USPTO_50K and Yes Monte Carlo tree searching USPTO_MIT) AiZynthFinder[159] Retrosynthetic 2020 Monte Carlo tree search guided by an USPTO Yes artificial neural network policy

AIndicates the year that a publication, conference proceeding, or press release explicitly discussing the program appeared. It is clear both from previous publications by the same authors, and the sheer amount of time that developing these programs takes, that they were under development for some time before this. For instance, SYNCHEM, first disclosed in 1977, was already under development at the time that LHASA was announced in 1969. Similarly, the origins of Chematica/Synthia, originally published in 2013, are traced back to an MIT bridge club in 2001.[35a] It is perhaps telling of the speed at which the field moves that even in their initial publication on SYNCHEM the authors note they have developed a ‘disaffection’ for SYNCHEM and have abandoned its development in favour of its successor, SYNCHEM2.[13a] 310 C. M. Williams and M. A. Dallaston intelligence.[132] Whether this move was a software bug or designed to be interactive to allow the chemist to choose which deliberately executed (and indeed whether a computer can be branches of the synthesis tree were worth exploring based on said to ‘deliberately’ do anything) remains up for debate. Deep their own knowledge, favoured reactions, and experience in the Blue relied on human-programmed rules and strategies, and laboratory. This concept of being able to explore a synthesis tree could not learn new tactics simply by observing other players has been common to nearly all programs developed since, during the game. In this way, Deep Blue was analogous to the although the tree has sprouted considerably more branches early retrosynthesis programs such as LHASA and can be (Fig. 4). described as a ‘brute-force’ player, defeating opponents through Even though the methods employed by the programs can be sheer knowledge and ability to quickly evaluate every known (to described by umbrella terms (e.g. hand-coded rule set, machine the computer) option. More recently, board game-playing com- learning, a combination of both), often the authors of a program puters have used deep neural networks and are capable of not would select an overarching approach that differentiated it from only playing the games, but also of independently developing other contemporary programs. For example, SYNCHEM was new strategies.[133] These neural network computers display a unusual in that it first considered only the carbon skeleton of the game style described as more ‘human’ and ‘elegant’ than the target molecule and systematically broke C–C bonds to find the earlier computers such as Deep Blue, and are regularly used as most efficient (convergent) route from a database of starting training tools by today’s competitors. There is also a great deal materials; once starting materials were found, SYNCHEM gen- of research into using neural networks for chemical synthesis erated intermediates by adding all possible functionalities to the and retrosynthesis (see below). While neural networks show carbon skeletons and mechanistically testing the reactions and great promise for this type of problem, it is important to note that deleting non-viable ones, whereas other programs would start they are not without their issues, often being described as from considering the target molecule as a whole.[137] For ‘brittle’, i.e. incredibly strong until they are faced with some- forward prediction programs, EROS considered reactions as thing unknown, carelessly entered, or deliberately edited, when bond- and electron-shifting processes, and then implemented they can break in spectacular fashion.[134] evaluation processes to select only chemically viable reactions, Historically, two general methods of building computer- whereas many other prediction programs focus on functional aided retrosynthesis programs have emerged: (1) human- groups and their transformations as the basis of chemical coded rules combined with a heuristic search; and (2) machine reactions.[138] learning techniques that automatically generate rules from a Although it has been proposed that the synthesis of long- provided reaction database. Human-coded rule programs domi- ifolene (222) was Corey’s inspiration for LHASA, and likely the nated up until the mid-2000s, starting with Corey’s pioneering birthplace of the term retrosynthesis, it was not subjected to work in the late 1960s. OCSS (Organic Chemical Simulation of LHASA.[67] That said, it did form its ground rules.[1c] The Synthesis),[12a] which evolved to become LHASA (Logic and synthetic approach also utilised as the key step the santonic acid Heuristics Applied to Synthetic Analysis),[135] is considered the degradation performed by Woodward (see Scheme 11 first true computer-assisted retrosynthesis planning tool. It was above).[67] Therefore, based on this premise, it is worthwhile

OH (a)Structure index (b) 1. 1 11. 26. 27. 33. OH OH OH 16. 17. 32 31 37 39 40 42 43 49

11 26 27 34. 35. 55 5160 61 62 63

59 65 66 67 70 OH OH OH OH

16 17 32 31

X O O

34 35 33

39 37 40 42 43 49

55 60 51 61 OH62 63

OH 59 65 66O 67 70

Fig. 4. (a) A synthesis tree as drawn by LHASA in 1969 with the target compound at the top. From E. J. Corey and W. T. Wipke.[12a] Reprinted with permission from AAAS (The American Association for the Advancement of Science). (b) A synthesis tree as drawn by Chematica in 2015 with the target compound at the centre. Reprinted from T. Klucznik et al. with permission.[136] Copyright 2018, Elsevier. The Future of Retrosynthesis 311 discussing the longifolene synthesis in detail. The initial retro- Starting from the Wieland–Miescher ketone (229), the synthetic manoeuvre consisted of introducing carbonyls adja- A-ring carbonyl was protected as a ketal (230), which enabled cent to the gem-dimethyl group, and in place of the exocyclic the enone to be chain-extended (i.e. 231), thus setting the stage double bond (i.e. 223). These key carbonyl insertions provided for a pinacol rearrangement via diol 232. The six- to seven- retro-conjugate addition synthons to be formulated, which membered ring enlargement was the first key hurdle in the resulted in the enone (224) being proposed. Rationalising the synthetic approach, and although proceeding smoothly in the three-dimensional structure in two dimensions reveals 225, end, it did require activating conditions to promote the pinacol which could be derived from the Wieland–Miescher ketone rearrangement and afford the desired ring system (i.e. 234), (226), accessed via a Robinson annulation of methylcyclohex- albeit a different isomer to that proposed in the retrosynthesis andione (227) with methylvinylketone (228)(Scheme 22).[2b] (i.e. 225). The second key hurdle followed directly after in that conditions were required to mediate cage bicyclic formation. This was eventually achieved using triethylamine at high temperature, and only produced material (i.e. 223) in the order O O of 12 % yield. Having surmounted these hurdles, the remaining part of the synthetic campaign amounted to a single skeleton O O embellishment, and several FGIs, for example, completion of H the gem-dimethyl function (i.e. 235), followed by removal of the 222 223 224 adjacent ketone to give 237 via reduction of a dithiane (i.e. 236). Final oxidation of alcohol 237 and subsequent methylenation O O O over two steps gave the long-awaited and seminal target longifolene (222)[160] (Scheme 23). + O Interestingly, the Corey group also spawned many of the next O O O 228 226 H generation of programs when those that worked on OCSS and 227 225 LHASA moved on to their own independent careers. Jorgensen, whoworkedontheLHASA project as a graduate student in Corey’s Scheme 22. Retrosynthetic analysis of longifolene 222. laboratory, went on to develop CAMEO (Computer-Assisted

O (CH OH) 2 2 O O pTsOH O EtPPh3Br / nBuLi O Δ C6H6 / Et2O / THF O 66 % O r.t. to Δ 229 230 96 % 231

i) OsO4 / pyr / Et2O –20°C to r.t. ii) pyr / NaHSO3 H2O / 5–10°C

O O 1) pTsCl / pyr O O HCl / EtOH O DCM / 0–3°C O 90 % O 2) LiClO4 / CaCO3 THF / 50°C 234 HO 233 48 %, 3 steps 232 OH

NEt3 12 % (CH2OH)2 225°C 1) LiAlH / O 4 O S Et O MePh3Na 2 MeI HS SH S Δ (quant.) O O O Dioxane BF ·Et O 3 2 2) N2H4 / Na 58 % 72 % (CH OH) 223 235 236 2 2 195°C quant.

Δ 1) MeLi / Et2O / 93 % i) AcOH / CrO3 OH O 2) SOCl / CCl3F ii) MnSO4 / 55°C pyr / 0°C 82 % Longifolene 88 % 238 237 (222)

Scheme 23. Corey’s synthesis of longifolene. 312 C. M. Williams and M. A. Dallaston

Mechanistic Evaluation of Organic Reactions).[139,161] Instead Chiral Pool of retrosynthetic analysis, this was a forward prediction pro- The chiral pool, which has already been mentioned above gram, which relied on human-programmed mechanistic rules to (see for example Scheme 17), is the collection of optically predict the outcome of a reaction. In total, 26 papers detailing the active starting materials that are available from natural sources, development and capabilities of CAMEO were published; the [162] such as amino acids, carbohydrates, and terpenes. They are last appeared in 1995. It seems that ceasing work on highly utilised in synthesis, because of their exceptional enan- CAMEO was a conscious though difficult decision for Jorgen- [163] tiopurity, and thus can be incorporated into the final structure of sen, as he discussed in his autobiography in 2015. Wipke the target or as chiral auxiliaries. Furthermore, the chiral pool also worked on OCSS as a postdoctoral researcher and went on [141] [13b] presents an interesting opportunity for the use of computers to to develop both SST and SECS. It is likely that many recognise structural relationships between targets and starting early programs were abandoned because upkeep and expansion materials. of the rule set became too time-consuming, and the results were As a prime example, the chiral pool approach was used by not promising enough to warrant the commitment. Corey for the synthesis of helminthosporal (239).[167] Starting Since the 2000s, advances in machine learning and com- with S-(þ)-carvone (240), a natural product readily available puting power have opened the door to more complex machine from caraway seeds, provided both a chiral centre and the learning approaches. In terms of synthetic chemistry, three isopropyl side chain, which inserted the correct configuration learning modes are used: (1) supervised learning, where both of the isopropyl group in the target. In brief, hydrogenation of the input (reactants, reagents, conditions) and output 240 gave the saturated ketone (241), which was converted into a (products, yield) are provided to the computer in the form of sacrificial 1,3-dicarbonyl (i.e. 242) to induce smooth albeit slow a training set; (2) unsupervised learning, in which only the conjugate addition to methylvinylketone (228). Tricarbonyl 243 input data are provided to the computer; and (3) reinforcement was then decarbonylated (i.e. 244), and subsequently cyclised to learning, where the computer aims to find the optimal path [164] the bicyclo[3.3.1] system (i.e. 245) through treatment with the towards the goal and is given a reward for each step. The Lewis acid boron trifluoride. Wittig-mediated chain extension method or mode of learning is independent of the end-goal of the installed the masked aldehyde (246), which underwent a trans- program, i.e. a supervised learning mode can be utilised in a protection (i.e. direct interconversion of protecting groups[168]) retrosynthesis program, a forward prediction program, or any with ethylene glycol to give the ketal (247) (note no change in other end-goal. oxidation state converting an enol ether to a ketal). Dihydrox- In machine learning, the programs are trained using data- ylation gave the diol (248), which was oxidatively cleaved to bases of chemical reactions, and the quality of data (e.g. give the dicarbonyl (249) as the key step in setting up the ring- experimental details, stereochemistry, regioselective contraction strategy towards the bicyclo[3.2.1] system, the latter information) varies greatly between different databases. Often being achieved via a Claisen condensation affording the cage the information is extracted automatically and this can result in bicyclic enal (250). The final step entailed a simple acid- errors or missed information, which would have otherwise been mediated deprotection to unmask the remaining aldehyde in identified by a human, especially when the information is being helminthosporal (239)(Scheme 24). extracted from prose (written experimental procedures, patents) Such a synthesis is described as taking a chiron approach to rather than drawn reaction schemes (although drawn reaction retrosynthetic planning. Often, however, the relationship schemes often omit crucial information such as temperature and between the structure of the target and the chiral pool is not time).[155b] In a recent interview about a new collaboration, MIT immediately obvious, and in the case of sugars can result in researchers stated that they hoped the quality of data in the CAS much of the initial functionality being removed (e.g. to access (Chemical Abstracts Service) database would improve the the carbon framework) to achieve synthesis of the target (e.g. predictive power of their machine learning algorithm that has [169] previously only been trained on freely available datasets.[165] leukotriene B5 (251) from tri-O-acetyl-D-glucal (252) ). In While it seems intuitive that an algorithm trained on data rich in, the helminthosporal example, a substantial clue to considering a chiron approach, beyond the importation of stereochemistry, for example, stereochemical information would be better at þ predicting stereochemical outcomes of reactions, there is some was the isopropyl group and the cyclohexane moiety that S-( )- evidence that the dataset an algorithm is trained on may have carvone provides. Furthermore, when the hydrogenated material little effect on the quality of its predictions.[166] (i.e. 241) is redrawn in the chair configuration, the relationship A final consideration for both industry and academic between the starting material and target is more apparent, and chemists is that confidentiality and intellectual property is of now lends more weight to the adopted retrosynthetic design utmost importance. Many of the browser-based free programs (Scheme 25). Actually, the term chiron (chiral synthon) was coined in (e.g. IBM RXN) explicitly state that any information submitted [170] through their website (i.e. into the retrosynthesis program) is the early 1980s by Hanessian, who also developed a computer program specifically to search for chiral starting visible to the developers. IBM RXN go on to state that by [140a] providing information, the user is granting IBM unrestricted materials given a target molecule. CHIRON was an inter- use rights.[155] Some commercial services (e.g. SpayaAI) also active program that projected different perspectives of a make the search input data visible to the developers.[158] Others drawn molecule, allowing the chemist to ‘see’ the compound (e.g. ICSynth) can be installed locally and therefore the data are in new ways. The program also mapped stereochemical infor- only visible to the user.[150] For these subscription services, mation onto the target and searched its hand-assembled starting there is also the issue of what happens to the data on termination material database for materials with maximal overlap in func- of the licence or agreement. As the capabilities and availability tional and stereochemical features. The aim was to find starting of computer-assisted synthesis programs expand, the status of materials that would require the least transformation to reach the the ownership of the input and output data will continue to product, specifically focussing on the carbon framework and evolve. rings. The Future of Retrosynthesis 313

O O OO O H2 NaOCH3 NEt3 OH + O r.t. Pd-Al2O3 HCO2Et O 228 3 days EtOH 241 C6H6, 242 243 S-(+)-Carvone 240 ( ) EtOH K2CO3

O (CH2OH)2 PPh3=CHOMe BF3 O O TsOH DMSO CH2Cl2 H H H O C H , OMe H O 247 6 6 246 245 244

OsO4 Pyr

H HO O H O OH O O Pb(OAc) H SO 4 NaOH 2 4 H O O O AcOH, C6H6 EtOH THF / H O H H O H HO H HO HHO 2 248 249 250 Helminthosporal (239)

Scheme 24. Synthesis of helminthosporal 239 from S-(þ)-carvone 240.

(a) O CO2H AcO HO HO AcO OAc Leukotriene B5 (251) Triacetyl-D-glucal (252)

(b) OO OHC

CHO O H H 241 241 240 239

Scheme 25. (a) Leukotriene B5 (251) chiral pool synthesis from tri-O-acetyl-D-glucal (252); (b) S-(þ)-carvone (240) and its structural relationship to helminthosporal (239).

Currently Available Programs early now-abandoned programs that relied on this type of Reaction Predictor,[151] Seq2Seq,[154] IBM RXN for Chemis- database, will need to be kept up to date for the program to try,[155] LillyMol,[157] SpayaAI,[158] AutoSynRoute,[12e] and continue to be of use as new chemistry is discovered. Second, AiZynthFinder[159] are all currently freely available, either as Synthia has a library of algorithms that it can draw on. The browser-based applications or as code downloadable from algorithms can, for example, augment the reaction rules to GitHub. They all rely on machine learning methods for training, improve regio- and stereo-selectivity predictions, allow the and do not include any hand-coded reaction rules. ICSynth,a program to plan over several steps, or combine reactions that commercial machine learning-based program, has seen use in proceed under the same conditions (i.e. tandem reactions) into the pharmaceutical industry for route design.[150] one step. These algorithms set it apart from other programs, Arguably the most advanced computer-aided synthesis tool which can only plan one step at a time. today is Synthia (formerly Chematica) developed by Grzybow- This hybrid hand-coded-rules-and-algorithm approach ski et al. and recently licenced by Merck. From its conceptual means that Synthia approaches retrosynthetic problems in a beginnings in 2001 at the MIT Bridge Club[35a] to being seemingly more ‘human’ way than other programs, which can validated by in-laboratory results in 2018, the road to this only apply hand-coded rules strictly in the way they were autonomous synthesis planning tool was certainly a long written, or may miss an important reaction that is not covered one.[136] The technical details of how Synthia works are covered in the database it was trained on. Synthia has probably the largest in detail elsewhere, and so here only a few differences and chemical reaction rule knowledge set of any program thus far similarities between this program and others that have come (both as a consequence of the effort put in to coding the rules and before will be mentioned. by virtue of there simply having been more chemistry known First, despite the rapid advances and interest in machine now than in the past). learning of late, at the core of Synthia’s workings is a massive In 2018, verification of Chematica’s capabilities was published, with the in-laboratory confirmation of several routes database of over 50000 human-coded reaction rules. These rules [136a] were compiled by chemists over many years and, much like the devised by the program (Fig. 5). The Chematica-devised 314 C. M. Williams and M. A. Dallaston

N O O N S N N O N O S N N OH HN N N SO2 Cl Cl O HN 253 254 255 MS: 8 steps, ~1 % yield MS: 'too risky' to prioritise MS: 7 steps, <10 % CH: 4/5 steps, 40 % yield CH: 7 steps, 3.2 % yield CH: 4 steps, 20–22 %

OH CF3 O O OMe HN NH HN O O

N N OH S OH O N OH 256 257 258 MS: 1 % yield MS: Literature procedure failed Not previously synthesised CH: 6 steps, 60 %yield CH: 5 steps, 20 % yield CH: 72 % yield (88 % ee)

S N(nBu)2 Avoiding N patented routes N O (41 % yield) O N Only known route patented CH: 40 % yield O CH: 55 % yield nBu H N (2.5× higher than patented) O OH HN O Ms H 259 260

Fig. 5. Pharmaceutically relevant targets synthesised via routes developed using Chematica (CH) and comparison with the previously published routes and/or MilliporeSigma (MS) routes. syntheses all provided improvements on the literature proce- Is the Target Real? dures in terms of yield and/or number of steps (253–257). In two Both natural and synthetic products are regularly assigned examples (i.e. 259 and 260), Chematica was used to specifically incorrect structures,[173] and it is often not until practitioners avoid patented routes and provided comparable or significantly attempt their synthesis for the first time (e.g. trunkamide A improved yield. One was a natural product (i.e. 258) that had not (261)),[174] or repeat a synthesis (e.g. tetrahydrooxazepine previously been synthesised. This verification clearly showed (262)),[175] that the error is detected and corrected. Com- that there has been great improvement in the synthesis planning pounding this problem, in some cases even achieving a synthesis capabilities of computer programs over the last 50 years; how- of the proposed target may not offer any additional information ever, the commercial aspect of the more capable programs on the proposed structure (e.g. afzeliindanone (263)).[176] (Synthia and ICSynth) means that their impact on the wider Another dimension to this problem is that some structures are research community will be limited. Synthia also risks encoun- extremely difficult to solve (e.g. EBC-232 (264)),[177] or an tering the same fate as earlier hand-coded programs, abandoned outright solution is not possible (e.g. a flat structure lacking owing to the time required and difficulty in keeping them up to relative and/or absolute stereochemistry). Therefore, the ques- date. Additionally, there is a great amount of research being [155b] tion in this context is whether AI development will in the future carried out in the area of automatic data extraction, neural attempt to verify the input target structure has been correctly networks,[171] natural language sequence to sequence proces- [129d,154] [172] elucidated, which is especially important for those suspicious sing, and machine learning, which negates the need structure proposals that defy physical organic concepts such as for manual extraction of data and updating reaction rules. aromaticity (e.g. 265)[178] and Bredt’s rule (e.g. 266)[179,180] (Fig. 6). Indeed, when describing an earlier incarnation of Discussion Chematica (pre 2016), Grzybowski et al. stated that ‘In full Having canvassed both humanistic and computer-assisted disregard of Bredt’s rules, it installed double bonds at bridge- retrosynthetic methods and practices in detail, it is impor- head atoms’, and described this as a ‘creative’ and ‘non-existent tant that with this knowledge, questions be posed and con- or improbable’ structural feature.[152a] The structures synthe- sidered to better understand a probable direction this art form sised via Chematica suggested routes in 2018 (Fig. 5) do not will take in the future, i.e. in the context of a computation era contain rare or unusually strained motifs, so it remains to be seen attempting to revolutionise the discipline of chemical syn- how Synthia would deal with such a structure now. thesis through both retrosynthetic design and forward syn- It is certainly conceivable that AI will indeed adopt an thetic planning. elucidation aspect, because it is currently possible through The Future of Retrosynthesis 315

Ph O S Not the natural O product N N N H O O HN MeO O N O NH H N HN HO O O O Tetrahydrooxazepine Afzeliindanone O (262) (263)

Trunkamide A OMe OH O (261) HO O OH O H O HO H H S HO OH N Neoveratrenone (266) NH OOH N H EBC-232 (264) 265

Fig. 6. Misassignments and elucidation errors. modern Computer-Assisted Structural Elucidation [177,181] OH (CASE) to screen target structures for validity. More Revised O importantly, this feature can be undertaken rapidly with current O [182] O O density functional theory (DFT) and associated methods to OH give surprisingly accurate re-assignments, for example, the Botryosphaerihydrofuran 268 (267) proposed structure for botryosphaerihydrofuran (267) and its d rmsd (d ) > 10 ppm rmsd ( C) = 1.14 ppm revised structure 268 (Fig. 7).[183] Assigning absolute stereo- C chemistry is a developing possibility with such methods.[184] Fig. 7. CASE reassignment using DU8þ. rmsd ¼ root mean square With that said, there are risks associated with automated deviation. NMR-based structural elucidation in terms of DFT energy miscalculations and subsequent erroneous consequences.[185]

Will AI Consider Model Systems? Model systems (or model studies) in organic synthesis are N N loosely defined as testing a ‘key step’ in a proposed synthesis H H without probing the real system. Such studies have been widely H OH N used, and in some cases with great success, for example, D establishing whether ring-closing metathesis was a viable option N for the construction of the D ring seen in manzamine A (269) H (Fig. 8).[186] However, the general view of the community is that a ‘model is a model, and real is real’,[7] meaning that the model Manzamine A system may or may not work, and that outcome is not a true (269) reflection of whether the proposed transformation would work (or not) in the real system. Therefore, it could be justified that Fig. 8. Manzamine A (269) with the D ring highlighted in blue. model systems be deemed obsolete if the practitioner has embraced full AI assistance. depends solely on the functional group appended to the parent Will AI Discount Fundamental Molecules? skeleton (Fig. 9). For example, iodinated systems (e.g. 271 and Fundamental molecules, such as strained hydrocarbons (e.g. 272) are quite sensitive, but not methyl esters (e.g. 273),[50b] cubane (54)[48] and [1.1.1]propellane (270)[187]), have fasci- although some carboxylic acids are strong candidates (e.g. nated theoretical and synthetic chemists since the discipline of 274).[50a] Hence, will AI search the literature for this know-how organic chemistry began. However, the issue with many of these or calculate strain energy before making such synthetic design systems is undesirable physical properties, for example, explo- proposals? With available literature, and the ability to calculate sivity and impact sensitivity. Therefore, will future AI retro- strain energy and impact sensitivity,[50a] it should be feasible to synthesis and synthetic planning programs take into account introduce these features. Information such as impact sensitivity, these details, which are a potential safety concern if not handled however, is often only reported in the written text and has no with practical knowledge and expertise? Cubanes and bicyclo standardised reporting protocol, meaning that this information is [1.1.1]pentanes are such examples, but their impact sensitivity very difficult to automatically extract from the literature. 316 C. M. Williams and M. A. Dallaston

intermediate or step, and this information came from the hand-coded reaction rules. It has recently been reported that 54 270 the latest version of Synthia will now explicitly include all protection and deprotection steps in the proposed syntheses, so I [191] OH this is clearly still an area under development. At this point I in time, it is also reliant on human choice to decide on whether a 271 272 I synthetic step can actually be performed without the use of a O protecting group. There are numerous syntheses that have been O completed without the use of a single protection or deprotection, OH HO OMe and this is a preferred situation as it limits synthetic steps, is economical, and limits waste.[192] However, there is no denying O 274 273 that syntheses without protecting group deployment are difficult, and if the synthetic route to the target (or the target itself) is Fig. 9. Cubanes, [1.1.1]propellane (270), and 1,3-diiodobicyclo[1.1.1] not conducive to it, then the effort involved in attempting to pentane (272) as fundamental molecule examples. achieve this goal may end up using many more resources (e.g. labour) than adopting a protecting group or two. That is, protecting group-free synthesis is an admirable goal, but must Overall, however, these intriguing molecule types are playing an be viewed in realistic terms. increasingly important role in bioactive molecule discovery,[51] and thus cannot be eliminated from consideration. Can Domino and Cascade Reactions be Readily Predicted? Domino and cascade reactions are always a welcome opportunity Will AI Learn from Practical Synthetic Problems? for the synthetic practitioner to implement because of numerous It is often the case with synthetic practice that retrosynthesis benefits, for example increased efficiency and waste minimisa- design is very much considered a rough plan for the forward tion of solvents, reagents, adsorbents, energy, and labour. synthesis, which in turn becomes a trial-and-error process to Therefore, such reactions facilitate ecologically and economi- achieve the target. This process has been elegantly described as cally favourable production and manufacture, and would [188] ‘dead ends and detours’. Given this aspect is an inherent greatly assist meeting Mulzer’s 25-step limit for any complex process of chemical synthesis, it is likely that AI will by default synthesis.[193] follow a similar path, but in collaboration with the practitioner to The terms domino, cascade, tandem, and one-pot are used determine what works in reality. This essential feedback loop apparently interchangeably in much of the literature, but they provides further information on unexpected failures arising are used to describe very different reaction procedures. A from the initial AI retrosynthetic plan and projected synthetic domino reaction, as defined by Tietze,[19] forms several bonds approach. This process will ultimately achieve the stated goal, in one sequence without isolating the intermediates, changing but it needs to be determined whether this approach is any dif- the reaction conditions, or adding reagents. Importantly, in a ferent for what is currently undertaken based on solely human domino reaction, each bond-forming reaction depends on the design. Efforts to evaluate such AI performance are ongoing; see previous step to set up the required functionality. A one-pot for example Fig. 5. reaction is similar in that it is a sequence of transformations Another aspect for consideration under this section is that of taking place without isolation of the intermediates; however, reaction yields. The main issue is that reported yields vary in further reagents are added to effect each step and reaction accuracy owing to a range of factors (e.g. scale of reaction). To conditions can be changed,[21] whereas a tandem reaction has complicate matters further, academic laboratories do not deter- been defined by Nicolaou et al. as two or more reactions mine yield values to the same level of accuracy being under- occurring on the same substrate and under the same conditions, [189] taken by industry. Hence, this can be considerably but essentially in isolation of one another.[21] Tandem is com- misleading when coding software for predictive forward syn- monly used to describe reactions that would be more accurately theses. This complication is unlikely to change anytime soon. defined as one-pot or sequential;[194] however, tandem reactions are not time-resolved and so this usage is not preferred as How will Protecting Groups be Interpreted? justified by Tietze.[19] According to Nicolaou et al., cascade The famous opening chapter to ‘Protecting Groups’ by can be used as an umbrella term for all domino, one-pot, and Kocien´ski, states up front ‘Death, Taxes and Protecting Groups’. tandem reactions,[21] Tietze argues that while cascade is often This very statement sets the stage for the necessary evil that used to describe domino reactions, it is ambiguous and the latter comes with performing chemical synthesis, which is only term is preferred.[19] exacerbated by an increasingly complex target that demands an As much as these reactions are highly desirable, their design extended synthetic route. Kocien´ski follows by introducing is by no means straightforward, and is generally associated with seven tactical considerations to define how effectively a pro- natural product syntheses due to inspiration provided by Mother tecting group should perform.[190] In essence, protection group Nature (see the Biomimetic Synthesis section above). In terms of chemistry is a very large and complicated component of per- AI, this places a significant demand on programming. Synthia forming chemical synthesis, and must be very well understood will identify reactions that can take place under the same for AI adoption. reaction conditions,[195] which addresses simultaneous or tan- Synthia is currently the only program that fully engages and dem reactions, using the definition of Nicolaou et al.[21] Tactical utilises the concept of protecting groups, because it is the only combinations, as coined by Corey,[1a] can be defined as strategic software package that considers strategies over more than one retrosynthetic transformation sequences, the first of which does synthetic step. Previous versions of Synthia indicated to the not immediately or obviously produce any benefit and often user when protecting groups were required for a particular increases molecular complexity, but sets the scene for the The Future of Retrosynthesis 317 second transformation, which results in molecular simplifica- molecular complexity had to be considered by the Chematica/ tion. The term can in theory be applied to any combination of Synthia team, and was quantified by the number of rings, the two transformations that fit the criteria; however, previously number of stereocentres, and the length of the SMILES string only ,500 such combinations had been explicitly identified in code of the molecule.[152b] Some targets, such as TaxolÒ (148) the literature and catalogued. Grzybowski et al. used Chematica are currently ‘too complex’ for Synthia, and the route search to discover ,4.85 million tactical combinations and subse- times out (i.e. exceeds the available RAM).[191] quently explicitly coded them into the program (along with In an effort to conceptualise retrosynthetic planning based on the ,500 ‘known’ tactical combinations).[152b] This is part of complexity, there are three general parameters that have been used the reason why Synthia can plan over multiple steps, but by the lead author over the years to categorise synthetic approaches currently only two-step combinations have been included.[152b] to targets. These three categories, which are essentially a ranking These tactical combinations are not necessarily suitable for system of synthetic difficulty, are best imagined as analogies to domino reactions, however, because there is no requirement proton NMR splitting patterns, which include first-order,[197] for them to preclude the isolation of intermediates. second-order,[198] and non-first-order[197] interpretations. Overall, it is possible that Synthia would identify a route that A good first-order example is the total synthesis of EBC-329 would, in practice, result in a domino process. However, it seems (275),[199] as it is mostly first-order (n þ 1) retrosynthetic bond that the user would have to manually review the retrosynthetic breaking. When further simplified by adopting a chiral pool tree and identify that each reaction step would take place approach, as suggested by the cyclopropyl-gem-dimethyl resi- sequentially and under the exact same reaction conditions, due, the conjugated enlactone is disconnected in the first unless the domino sequence was already known and was coded instance to give an enyne (276) as conjured from metal coupling. in as a reaction rule. Therefore, this implementation could be Further disconnection of the enyne to a methyl ketone (i.e. 277) some time away and is far from being perfected. is inspired by Wittig chemistry, as is the inspiration for the subsequent disconnect to the dicarbonyl (278). Two carbonyls Assessing Molecular Complexity distanced by six carbon units imply an oxidative ring cleavage In light of what we know about computer-assisted and AI strategy fitting carene (279)(Scheme 26). methods for devising retrosyntheses and route planning, there is Based on this first-order retrosynthesis, the forward synthesis a requirement for the computer to understand complexity. almost became pedestrian, albeit with one minor hurdle. Oxida- Complexity is two-fold in this context, although related: (1) tive cleavage performed best over two steps to give 278, which molecular or target complexity, and (2) synthetic design or route reacted smoothly with the phosphorane (280) to afford 277. complexity. Note that the molecular complexity of the target Some experimentation was required to procure the second may or may not be directly proportional to the number of steps olefination, but this was eventually achieved using Horner– required to synthesise that target. Therefore, should there be Wadsworth–Emmons (i.e. 281) methodology, although a mix- some consideration as to whether this impacts AI performance ture of E- and Z-isomers was obtained (i.e. 282). The final step in undertaking retrosynthetic design and predicting forward consisted of installing the g-alkylidenebutenolide function, syntheses more efficiently? Obviously, AI will inherently which was accomplished in one step via a copper-mediated attempt to provide the ‘best’ synthetic approach following reaction with iodoacid 283 (Scheme 26). HPLC purification consideration of the proposed retrosynthetic plan. then afforded the pure natural product, establishing the absolute How is complexity perceived? Although molecular com- stereochemistry in a total of seven steps (Scheme 27). plexity is difficult for the practitioner to define, it is generally Second-order considerations include multiple bond discon- intuitively understood. For example, taking into account the nections at once, which would equate to strategies consisting of target’s carbon skeleton, number of appendages, and functional skeletal rearrangements (e.g. Scheme 21), cascade reactions groups, stereochemistry and conformation,[196] one can quickly (e.g. Scheme 5), and cycloadditions (e.g. Scheme 13) as prime ascertain a level of molecular complexity, and subsequently examples in this category. synthetic difficulty. As mentioned above, the latter is not Non-first-order retrosynthesis (akin to aromatic ring- necessarily determined by molecular complexity, i.e. complex splitting patterns) encompasses irrational bond breaking and molecules can be readily synthesised, but it is all dependent on forming retrosynthetic manoeuvres that generally only become retrosynthetic planning. When coding tactical combinations, clear once viewed. A seminal contribution to this category was

O O

CO H CO2Me 2 276 EBC-329 (275)

H H O O O CO2Me 278 277 (+)-2-Carene (279)

Scheme 26. EBC-329 (275) retrosynthetic plan. 318 C. M. Williams and M. A. Dallaston

1) OsO / NMO 4 PPh3 Me CO / H O 280 H H 2 2 20°C O CO2Me CO2Me CH CI / 20°C 2) PhI(OAc)2 O 2782 2 O 277 20°C 80 % 279 49 % over 2 steps 2) TBAF 1) NaHMDS / THF AcOH / r.t. –78°C I O TMS O 3) NaOH THF / H O OH P 2 70°C 283 281 OiPr CO2H OiPr Cul / K CO 2 3 282 DMF / 55°C 63 % H over 3 steps

HPLC CO2H 31 % (+)-EBC-329 O (275)

Scheme 27. Asymmetric total synthesis of EBC-329 (275). the total synthesis of nominine (284) by Peese and Gin.[200] The retrosynthesis boldly aimed to build four bonds using a conver- OH gent dual-cycloaddition approach via key intermediate 285 N X (Scheme 28). N Nominine The elegant forward synthesis is immediately daring by not (284) 285 only starting with an unstable dimethyl acetal (i.e. 286), but performing an ortho-lithiation on a substrate where multiple Scheme 28. Gin’s retrosynthesis of nominine (284). positions could have in principle been lithiated. The product (288) was converted into a functionally dense azide (289), which was converged with a cyclohexylaldehyde (290) as obtained from enone 20 via triflate 291. The audacious Staudinger perspective? This aspect is difficult to judge because some reaction afforded the adduct (292) in high yield, which on chemical processes take decades to elucidate, although treatment with acid gave the desired zwitterion (293), setting not hindering their use and adoption by the community. The the stage for the first key 1,3-dipolar cycloaddition. Astound- Ley–Griffith oxidant, tetra-n-propylammonium perruthenate ingly, heating 293 at 1808C gave in one step the majority of the (TPAP), is a good example that was routinely used for mild and target skeleton (i.e. 294). FGI gave rise to the second cycloaddi- controlled oxidation of alcohols and many other substrates, but tion precursor, alkene 297, via intermediates 295 and 296. whose mechanism of oxidation was only recently elucidated, [202] Enamine formation derived from 297 facilitated the second having stood unknown for almost four decades. In the eyes cycloaddition to give 298 in 78 % yield. Installation of the of the computer, it is likely going to be a case that if it has lit- exocyclic double bond followed by allylic oxidation gave erature precedent as working, then it can be adopted irrespective nominine (284)(Scheme 29). This bold and exemplary case if the mechanism is understood or not. reduced a previously 40-step synthesis to just 15 steps. Given practitioners can intuitively categorise both forms of Synthetic Methodology complexity (molecular and synthetic), an in silico complexity From the perspective of AI coding, and current day AI perfor- index has been developed to rank both molecule and synthetic mance (e.g. Synthia), has sufficient methodology already been complexity,[201] which opens the door for AI adoption if a discovered? If the paradigm shifts introduced by ring-closing requirement is justified. The overall message being conveyed, metathesis[203] and C–H activation[10] are any indication, then however, is that complexity is much better understood by the methodological development must continue, whether by ratio- practitioner than the computer at this point in time. nal design or accidental discovery. However, what must not happen is the removal of old methodology from consideration as Mechanism new methodology is developed, i.e. this is not a case of Understanding the mechanism of a chemical process is a fun- replacement, but bolstering the repertoire. damental part of the practitioner’s toolbox, as it enables further discovery and innovation. However, is this process important for Will AI Develop a Chemical Mind or Brain Equivalent? retrosynthesis and forward synthesis if it merely comes down to Perhaps the most important, albeit emotionally confronting, bond breaking and bond forming from a graph theory or AI question for practitioners of synthetic chemistry is whether they The Future of Retrosynthesis 319

O i) nBuLi / THF 1) NaN3 OMe O Cl O N3 286 –23°C Acetone / r.t. O O MeO MeO ii) O 2) AcCl / MeOH OMe Cl OMe r.t. MeO NMeOMe 289 (94 % 287 288 (52 %) over 2 steps)

CN CHO i) Et2AlCN i) DIBALH PBu3 C6H6 / r.t. Toluene / 0°C NaBH(OAc)3 CH Cl / r.t. ii) Zn(CN) / C H 2 2 ii) TBAT / Tf2O 2 6 6 O OTf Pd(PPh ) / DMF C6H6 / r.t. 3 4 CN 20 81 % 291 60°C 290 (85 %)

OMe O THF / 180°C N TFA / H2O N H O 97 % O CH2Cl2 CN O 0°C CN MeO 293 (93 %) 292 (79 %)

O 1) NaBH4 / EtOH / r.t. i) DIBALH CN Δ CN O 2) SOCl2 / CH2Cl2 / O Toluene / 0°C

N 3) Bu3SnH / AIBN N 2) Ph3P=CH2 Δ 294 C6H6 / 295 THF / r.t. 63 % over 3 steps 82 %

i) Na / iPrOH NH THF / –78°C O O O MeOH ii) HCl / H O N N 2 N 60°C 97 % 298 (78 %) 297 296

1) Ph3P=CH2 OH THF / 70°C

2) SeO2 / tBuO2H N CH2Cl2 / r.t. Nominine 51 % over 2 steps (284)

Scheme 29. Gin’s total synthesis of nominine (276). will become obsolete owing to AI thinking and robotics (e.g. applicable to chemical synthesis programs.[209] Chematica was performing chemical reactions).[204] In fact, this is an area of recently subjected to the Feigenbaum or Turing test by having debate for humankind in general and across all facets of life. 18 experts grade the likelihood of 40 (20 from the literature and As early as 1950, the question ‘Can machines think?’ was 20 from Chematica) synthetic routes to be human- or computer- proposed by Turing, who also embarked on defining the term designed. Overall, the experts were unable to differentiate ‘think’.[205] Some 30 years later, Searle proposed essentially the between the human- and computer-designed routes, and even same question, ‘Could a machine think?’ but attempted to scored the computer-designed routes as being slightly more address the mind, brain, and program interface using the argu- ‘elegant’ than the human-designed routes.[191] ment of intentionality in human beings.[206] He then went on to With that said, all these considerations are philosophical differentiate the brain–mind relationship (dualism), often called debate, and are in the process of being translated, tested, and the mind–body problem, by considering the brain as a digital determined in silico. Therefore, it seems a long way off before computer, and the mind as a computer program.[207] More synthetic chemists undergo complete morphological change recently, Signorelli considered the idea of machines overcoming into scientific generalists, i.e. when a ‘scientist’ of the future humans and, if so, that they must be intrinsically related to acquires molecules (based on society drivers) using the push of a conscious machines.[208] button. The Turing test, originally called the imitation game by Turing, essentially asks whether an interrogator would be able to tell the difference between two participants, one of which is Conclusion human and one a machine.[205] Importantly, the Turing test There is no doubt that synthetic chemistry has been and will examines behaviour (output) rather than mechanism or working continue to be a substantial enabling foundation discipline for (process). There have been many variations of the Turing test, humankind. The past and present output of pharmaceuticals, and the subject matter expert or Feigenbaum test is particularly agrochemicals, and materials (e.g. plastics, microelectronics) is 320 C. M. Williams and M. A. Dallaston

a reflection of the magnitude of this discipline being deployed question regarding the proposed research program, which was not likely by industry.[210] answered satisfactorily. At the end of the lecture C.M.W. asked his host However, even though this branch of science is highly Professor Phil Parsons (now at Imperial College London), ‘Who was that?’, sophisticated, it is by no means trouble-free, as Cornforth to which he laughed and replied, ‘That was John Cornforth!’ Perhaps it was flagged back in 1993,[211] likely with Seebach’s 1990 proposal no surprise that Associate Professor Gareth Rowlands (now at Massey of ‘Organic synthesis – where now?’[212] front of mind. If we University, NZ) ended up getting the position. Lastly, the authors are most grateful to those that have provided feedback accept the answer to Seebach’s question is ‘everywhere’ and on this topic over the years and the ensuing manuscript. that Cornforth’s conjecture remains correct, then evolution of the discipline is inevitable, and in modern times that requires References integration of algorithmic methods to assist in prompting human [1] (a) E. J. Corey, X.-M. Cheng, The Logic of Chemical Synthesis 1995 decision-making. Therefore, with the emergence of programs (Wiley: New York, NY). such as Synthia, the real quandary is whether the art of organic (b) E. J. Corey, Chem. Soc. Rev. 1988, 17, 111. doi:10.1039/ synthesis will be maintained, i.e. will the syntheses be elegant, CS9881700111 and meet all social related desires of efficiency and environ- (c) E. J. Corey, Pure Appl. Chem. 1967, 14,19. doi:10.1351/ mental impact? Beyond that point, will computer-assisted PAC196714010019 methods and AI truly be revolutionary for the discipline, and [2] (a) The Nobel Prize, The Nobel Prize in Chemistry 1990. Nobel importantly should they be relied on completely as Judson Media AB 2020. Available at: https://www.nobelprize.org/prizes/ intimates?[213] chemistry/1990/summary (accessed 11 Jun 2020) (b) E. J. Corey, Angew. Chem. Int. Ed. Engl. 1991, 30, 455. ‘The GPS navigation device may render paper maps redun- doi:10.1002/ANIE.199104553 dant – but not the driver of the car.’[214] [3] (a) S. Warren, Designing Organic Syntheses: A Programmed Intro- duction to the Synthon Approach 1982 (Wiley: New York, NY). Lastly, the present article expresses certain thoughts and (b) S. Warren, P. Wyatt, Organic Synthesis: The Disconnection philosophies on retrosynthesis and forward synthetic planning Approach, 2nd Edn 2008 (John Wiley & Sons: New York, NY). that are designed to generate discussion among students and [4] M. B. Smith, J. Chem. Educ. 1990, 67, 848. doi:10.1021/ED067P848 colleagues alike. It is fully realised that this important and ever- [5] (a) J. Saunders, Top Drugs: Top Synthetic Routes 2000 (Oxford evolving area comes with many individual views and experi- University Press: Oxford). ences that constantly drive synthetic organic chemistry thinking, (b) T. P. Stockdale, C. M. Williams, Chem. Soc. Rev. 2015, 44, 7737. and Sir John Warcup Cornforth[215] is one of several iconic doi:10.1039/C4CS00477A individuals who have made significant contributions to the [6] K. C. Nicolaou, E. J. Sorensen, Classics in Total Synthesis: Targets, Strategies, Methods 1996 (Wiley-VCH: Weinheim). evolution of chemical synthesis. [7] K. C. Nicolaou, S. A. Snyder, Classics in Total Synthesis II: More Targets, Strategies, Methods 2003 (Wiley-VCH: Weinheim). Conflicts of Interest [8] K. C. Nicolaou, J. S. Chen, Classics in Total Synthesis III: Further The authors declare no conflicts of interest. Targets, Strategies, Methods 2011 (Wiley-VCH: Weinheim). [9] (a) D. G. Brown, J. Bostro¨m, J. Med. Chem. 2016, 59, 4443. doi:10.1021/ACS.JMEDCHEM.5B01409 Declaration of Funding (b) S. D. Roughley, A. M. Jordan, J. Med. Chem. 2011, 54, 3451. This research did not receive any specific funding. The Uni- doi:10.1021/JM200187Y versity of Queensland provided all the resources to enable the [10] (a) H. M. L. Davies, K. Liao, Nat. Rev. Chem. 2019, 3, 347. preparation of this manuscript. doi:10.1038/S41570-019-0099-X (b) Y. Wei, P. Hu, M. Zhang, W. Su, Chem. Rev. 2017, 117, 8864. doi:10.1021/ACS.CHEMREV.6B00516 Acknowledgements (c) J. He, M. Wasa, K. S. L. Chan, Q. Shao, J.-Q. Yu, Chem. Rev. C.M.W. is eternally grateful to the many bright and talented students and 2017, 117, 8754. doi:10.1021/ACS.CHEMREV.6B00622 post-doctoral researchers who have passed through his laboratories, bringing [11] D. Petzold, M. Giedyk, A. Chatterjee, B. Ko¨nig, Eur. J. Org. Chem. culture, philosophy, and humour to the Williams research group enterprise. 2020, 1193. doi:10.1002/EJOC.201901421 C.M.W. is profoundly grateful for financial and collaborative support over [12] (a) E. J. Corey, W. T. Wipke, Science 1969, 166, 178. doi:10.1126/ the last 20 years from the Australian Research Council (ARC), National SCIENCE.166.3902.178 Health and Medical Research Council (NHMRC), State and Federal Gov- (b) E. J. Corey, W. J. Howe, H. W. Orf, D. A. Pensak, G. Petersson, ernment departments (e.g. DPI, DAWE), Defence Science Technology J. Am. Chem. Soc. 1975, 97, 6116. doi:10.1021/JA00854A026 Group (DSTG), Commonwealth Scientific and Industrial Research Orga- (c) E. J. Corey, W. L. Jorgensen, J. Am. Chem. Soc. 1976, 98, 189. nisation (CSIRO), and local and international industry (e.g. EcoBiotics Ltd, doi:10.1021/JA00417A030 Bayer, Dow, DuPont). C.M.W. is eternally grateful for the many wonderful (d) E. J. Corey, A. K. Long, S. D. Rubenstein, Science 1985, 228, 408. scientific collaborations that have developed over time, and in turn allowed doi:10.1126/SCIENCE.3838594 his research to blossom in a multitude of unexpected directions. Such rela- (e) K. Lin, Y. Xu, J. Pei, L. Lai, Chem. Sci. 2020, 11, 3355. tionships with both academic and industrial collaborators are dearly cher- doi:10.1039/C9SC03666K ished. M.A.D gratefully acknowledges the Australian Government Research (f) P. Schwaller, R. Petraglia, V. Zullo, V. H. Nair, R. A. Training Program Scholarship. Haeuselmann, R. Pisoni, C. Bekas, A. Iuliano, T. Laino, Chem. Sci. C.M.W. is especially grateful to Emeritus Professor Curt Wentrup 2020, 11, 3316. doi:10.1039/C9SC05704H (University of Queensland) for the invitation to write an Australian Journal [13] (a) H. L. Gelernter, A. F. Sanders, D. L. Larsen, K. K. Agarwal, R. H. of Chemistry Sir John Cornforth review article. Beyond the invitation itself, Boivie, G. A. Spritzer, J. E. Searleman, Science 1977, 197, 1041. C.M.W. is especially honoured to contribute to this series having had the doi:10.1126/SCIENCE.197.4308.1041 fortunate opportunity to meet Sir John W. Cornforth Jr. during the course of (b) W. T. Wipke, G. I. Ouchi, S Krishnan, Artif. Intell. 1978, 11, 173. being interviewed for a Lectureship in 1999 at the University of Sussex, doi:10.1016/0004-3702(78)90016-4 Brighton, UK. Anecdotally, Sir John Cornforth sat in the front row of the (c) S. Szymkuc´, E. P. Gajewska, T. Klucznik, K. Molga, P. Dittwald, audience about 1 m away from the presenting interviewees, and at the end of M. Startek, M. Bajczyk, B. A. Grzybowski, Angew. Chem. Int. Ed. C.M.W’s lecture, asked (shouted owing to deafness) a stereochemistry 2016, 55, 5904. doi:10.1002/ANIE.201506101 The Future of Retrosynthesis 321

[14] J. W. Cornforth, R. H. Cornforth, K. K. Mathew, J. Chem. Soc. 1959, Crews, M. T. Crimmins, M. Csete, P. Dewapriya, J. M. Egan, M. J. 2539. doi:10.1039/JR9590002539 Garson, G. Genta-Jouve, W. H. Gerwick, H. Gross, M. K. Harper, P. [15] The carbon unit equation approach has been recently deployed to Hermanto, J. M. Hook, L. Hunter, D. Jeannerat, N.-Y. Ji, T. A. construct an all-trans-polyprenol with 100 carbons; see: Y. Totsuka, Johnson, D. G. I. Kingston, H. Koshino, H.-W. Lee, G. Lewin, J. Li, Y. Yasuno, T. Shinada, Chem. Lett. 2019, 48, 491. doi:10.1246/CL. R. G. Linington, M. Liu, K. L. McPhail, T. F. Molinski, B. S. Moore, 190046 J.-W. Nam, R. P. Neupane, M. Niemitz, J.-M. Nuzillard, N. H. [16] The non-traditional version of the retrosynthesis arrow (i.e. a closed Oberlies, F. M. M. Ocampos, G. Pan, R. J. Quinn, D. S. Reddy, J.- in forward arrow) is frowned on by the authors and is never utilised. H. Renault, J. Rivera-Cha´vez, W. Robien, C. M. Saunders, T. J. The authors believe that this arrow has introduced confusion to the Schmidt, C. Seger, B. Shen, C. Steinbeck, H. Stuppner, S. Sturm, O. discipline because it indicates a forward direction rather than reverse. Taglialatela-Scafati, D. J. Tantillo, R. Verpoorte, B.-G. Wang, C. M. The authors surmise that the introduction of the non-traditional arrow Williams, P. G. Williams, J. Wist, J.-M. Yue, C. Zhang, Z. Xu, C. arose in the first Classics in Total Synthesis book series.[6] Simmler, D. C. Lankin, J. Bisson, G. F. Pauli, Nat. Prod. Rep. 2019, [17] (a) For a detailed overview of retrosynthesis, see: M. B. Smith, 36,35. doi:10.1039/C7NP00064B Organic Synthesis 4th Edn 2016 (Academic Press) and preceding [34] (a) J. Y. W. Mak, C. M. Williams, Chem. Commun. 2012, 287. Edns 1–3. doi:10.1039/C1CC15995J (b) K. C. Nicolaou, C. R. H. Hale, Natl. Sci. Rev. 2014, 1, 233. (b) J. Y. W. Mak, C. M. Williams, Eur. J. Org. Chem. 2012, 2001. doi:10.1093/NSR/NWU001 doi:10.1002/EJOC.201101796 [18] (a) M. C. de la Torre, M. A. Sierra, Angew. Chem. Int. Ed. Engl. 2004, [35] (a) A. P.-J. Chen, C. C. Mu¨ller, H. M. Cooper, C. M. Williams, Org. 43, 160. doi:10.1002/ANIE.200200545 Lett. 2009, 11, 3758. doi:10.1021/OL901406J (b) M. Razzak, J. K. De Brabander, Nat. Chem. Biol. 2011, 7, 865. (b) A. P.-J. Chen, C. C. Mu¨ller, H. M. Cooper, C. M. Williams, doi:10.1038/NCHEMBIO.709 Tetrahedron 2010, 66, 6842. doi:10.1016/J.TET.2010.06.056 [19] L. F. Tietze, Chem. Rev. 1996, 96, 115. doi:10.1021/CR950027E [36] (a) M. J. Gallen, C. M. Williams, Org. Lett. 2008, 10, 713. [20] J. C. Wasilke, S. J. Obrey, R. T. Baker, G. C. Bazan, Chem. Rev. 2005, doi:10.1021/OL702827X 105, 1001. doi:10.1021/CR020018N (b) M. J. Gallen, C. M. Williams, Eur. J. Org. Chem. 2008, 4697. [21] (a) K. C. Nicolaou, D. J. Edmonds, P. G. Bulger, Angew. Chem. Int. doi:10.1002/EJOC.200800574 Ed. 2006, 45, 7134. doi:10.1002/ANIE.200601872 [37] (a) G. A. Strohmeier, H. Pichler, O. May, M. Gruber-Khadjawi, (b) K. C. Nicolaou, J. S. Chen, Chem. Soc. Rev. 2009, 38, 2993. Chem. Rev. 2011, 111, 4141. doi:10.1021/CR100386U doi:10.1039/B903290H (b) H.-P. Meyer, E. Eichhorn, S. Hanlon, S. Lu¨tz, M. Schu¨rmann, R. [22] J. Y. W. Mak, C. M. Williams, Nat. Prod. Rep. 2012, 29, 440. Wohlgemuthf, R. Coppolecchiag, Catal. Sci. Technol. 2013, 3,29. doi:10.1039/C2NP00067A doi:10.1039/C2CY20350B [23] K. Kawazu, Agric. Biol. Chem. 1980, 44, 1367. doi:10.1271/ (c) M. Winkler, M. Geier, S. P. Hanlon, B. Nidetzky, A. Glieder, BBB1961.44.1367 Angew. Chem. Int. Ed. 2018, 57, 13406. doi:10.1002/ANIE. [24] M. Kubo, M. Nakai, K. Harada, Y. Fukuyama, Tetrahedron 2019, 75, 201800678 2379 and references therein. doi:10.1016/J.TET.2019.02.049 [38] L. A. Baker, C. M. Williams, P. V. Bernhardt, G. W. Yanik, Tetra- [25] Y. Fukuyama, M. Kubo, H. Minami, H. Yuasa, A. Matsuo, T. Fujii, hedron 2006, 62, 7355. doi:10.1016/J.TET.2006.05.030 M. Morisaki, K. Harada, Chem. Pharm. Bull. 2005, 53,72. [39] P. A. Wender, Nat. Prod. Rep. 2014, 31, 433. doi:10.1039/ doi:10.1248/CPB.53.72 C4NP00013G [26] M. G. Constantino, V. L. Ju´nior, G. V. Jose´ Da Silva, Molecules 2002, [40] P. A. Wender, V. A. Verma, T. J. Paxton, T. H. Pillow, Acc. Chem. 7, 456. doi:10.3390/70500456 Res. 2008, 41,40. doi:10.1021/AR700155P [27] B. D. Schwartz, D. P. Tilly, R. Heim, S. Wiedemann, C. M. Williams, [41] P. E. Eaton, T. W. Cole, J. Am. Chem. Soc. 1964, 86, 962. P. V. Bernhardt, Eur. J. Org. Chem. 2006, 3181. doi:10.1002/EJOC. doi:10.1021/JA01059A072 200600246 [42] P. E. Eaton, T. W. Cole, J. Am. Chem. Soc. 1964, 86, 3157. [28] M. J. Gallen, R. Goumont, T. Clark, F. Terrier, C. M. Williams, doi:10.1021/JA01069A041 Angew. Chem. Int. Ed. 2006, 45, 2929. doi:10.1002/ANIE. [43] M. A. Ogliaruso, M. G. Romanelli, E. I. Becker, Chem. Rev. 1965, 65, 200504156 261. doi:10.1021/CR60235A001 [29] (a) A. Porzelle, C. M. Williams, B. D. Schwartz, I. R. Gentle, Synlett [44] K. Hafner, K. Goliasch, Chem. Ber. 1961, 94, 2909. doi:10.1002/ 2005, 2923. doi:10.1055/S-2005-921890 CBER.19610941112 (b) B. D. Schwartz, A. Porzelle, K. S. Jack, J. M. Faber, I. R. Gentle, [45] E. T. McBee, R. K. Meyers, J. Am. Chem. Soc. 1955, 77,88. C. M. Williams, Adv. Synth. Catal. 2009, 351, 1148. doi:10.1002/ doi:10.1021/JA01606A028 ADSC.200800739 [46] S. D. Houston, B. A. Chalmers, G. P. Savage, C. M. Williams, Org. [30] W. A. Eger, R. L. Grange, H. Schill, R. Goumont, T. Clark, C. M. Biomol. Chem. 2019, 17, 1067. doi:10.1039/C8OB02959H Williams, Eur. J. Org. Chem. 2011, 2548. doi:10.1002/EJOC. [47] (a) E. J. Ko, G. P. Savage, C. M. Williams, J. Tsanaktsidis, Org. Lett. 201001680 2011, 13, 1944. doi:10.1021/OL200290M [31] E. Winterfeldt, H. Bauermeister, H. Riechers, D. Schomburg, P. (b) J. Ho, J. Zheng, R. Meana-Pan˜eda, D. G. Truhlar, E. J. Ko, G. P. Washausen, Angew. Chem. Int. Ed. Engl. 1991, 30, 191. doi:10.1002/ Savage, C. M. Williams, M. L. Coote, J. Tsanaktsidis, J. Org. Chem. ANIE.199101911 2013, 78, 6677. doi:10.1021/JO400927Y [32] A. P.-J. Chen, C. M. Williams, Org. Lett. 2008, 10, 3441. (c) E. J. Ko, C. M. Williams, G. P. Savage, J. Tsanaktsidis, Org. doi:10.1021/OL801117E Synth. 2012, 89, 471. doi:10.15227/ORGSYN.089.0471 [33] Considering that the natural product isolation procedure used metha- [48] (a) P. E. Eaton, Angew. Chem., Int. Ed. Engl. 1992, 31, 1421. nol as the extracting solvent (often for weeks), there existed the [Angew. Chem. 1992, 104, 1447. doi:10.1002/ange.19921041105]. potential that the methylated systems were artefacts of isolation; see doi:10.1002/ANIE.199214211 for example: (a) R. J. Capon, Nat. Prod. Rep. 2020, 37,55. (b) K. F. Biegasiewicz, J. R. Griffiths, G. P. Savage, J. Tsanaktsidis, doi:10.1039/C9NP00013E R. Priefer, Chem. Rev. 2015, 115, 6719. doi:10.1021/CR500523X (b) L. A. Maslovskaya, A. I. Savchenko, V. A. Gordon, P. W. Reddell, [49] (a) M. J. Falkiner, S. W. Littler, K. J. McRae, G. P. Savage, J. C. J. Pierce, P. G. Parsons, C. M. Williams, Chem. – Eur. J. 2017, 23, Tsanaktsidis, Org. Process Res. Dev. 2013, 17, 1503. doi:10.1021/ 537. doi:10.1002/CHEM.201604674 OP400181G (c) J. B. McAlpine, S.-N. Chen, A. Kutateladze, J. B. MacMillan, G. (b) D. E. Collin, E. H. Jackman, N. Jouandon, W. Sun, M. E. Light, Appendino, A. Barison, M. A. Beniddir, M. W. Biavatti, S. Bluml, A. D. C. Harrowven, B. Linclau, Synthesis 2021, 53, 1307. doi:10.1055/ Boufridi, M. S. Butler, R. J. Capon, Y. H. Choi, D. Coppage, P. S-0040-1705964 322 C. M. Williams and M. A. Dallaston

[50] (a) M. A. Dallaston, J. S. Brusnahan, C. Wall, C. M. Williams, [72] A. C. Willis, P. D. O’Connor, W. C. Taylor, L. N. Mander, Aust. J. Chem. – Eur. J. 2019, 25, 8344. doi:10.1002/CHEM.201901086 Chem. 2006, 59, 629. doi:10.1071/CH06266 (b) M. A. Dallaston, S. D. Houston, C. M. Williams, Chem. – Eur. J. [73] L. N. Mander, A. C. Willis, A. J. Herlt, W. C. Taylor, Tetrahedron 2020, 26, 11966. doi:10.1002/CHEM.202001658 Lett. 2009, 50, 7089. doi:10.1016/J.TETLET.2009.10.019 [51] (a) B. A. Chalmers, H. Xing, S. D. Houston, C. Clark, S. Ghassabian, [74] T. A. Bradford, A. C. Willis, J. M. White, A. J. Herlt, W. C. Taylor, A. Kuo, B. Cao, A. Reitsma, C.-E. P. Murray, J. E. Stok, G. M. Boyle, L. N. Mander, Tetrahedron Lett. 2011, 52, 188. doi:10.1016/J.TET C. J. Pierce, S. W. Littler, D. A. Winkler, P. V. Bernhardt, C. Pasay, LET.2010.10.011 J. J. De Voss, J. McCarthy, P. G. Parsons, G. H. Walter, M. T. Smith, [75] P. Lan, A. J. Herlt, A. C. Willis, W. C. Taylor, L. N. Mander, ACS H. M. Cooper, S. K. Nilsson, J. Tsanaktsidis, G. P. Savage, C. M. Omega 2018, 3, 1912. doi:10.1021/ACSOMEGA.7B02065 Williams, Angew. Chem. Int. Ed. 2016, 55, 3580.[Angew. Chem. [76] S. V. Binns, P. J. Dunstan, G. M. Guise, G. M. Holder, A. F. Hollis, 2016, 128, 3644. doi:10.1002/ange.201510675]. doi:10.1002/ANIE. R. S. McCredie, J. T. Pinhey, R. H. Prager, M. Rasmussen, E. Ritchie, 201510675 W. Taylor, Aust. J. Chem. 1965, 18, 569. doi:10.1071/CH9650569 (b) T. A. Reekie, C. M. Williams, L. M. Rendina, M. Kassiou, J. Med. [77] L. N. Mander, E. Ritchie, W. C. Taylor, Aust. J. Chem. 1967, 20, 981. Chem. 2019, 62, 1078. doi:10.1021/ACS.JMEDCHEM.8B00888 doi:10.1071/CH9670981 [52] (a) L. Velluz, J. Valls, J. Mathieu, Angew. Chem. Int. Ed. 1967, 6, 778. [78] L. N. Mander, R. H. Prager, M. Rasmussen, E. Ritchie, W. C. Taylor, [Angew. Chem. 1967, 79, 774. doi:10.1002/ange.19670791706]. Aust. J. Chem. 1967, 20, 1473. doi:10.1071/CH9671473 doi:10.1002/ANIE.196707781 [79] L. N. Mander, R. H. Prager, M. Rasmussen, E. Ritchie, W. C. Taylor, (b) D. Urabe, T. Asaba, M. Inoue, Chem. Rev. 2015, 115, 9207. Aust. J. Chem. 1967, 20, 1705. doi:10.1071/CH9671705 doi:10.1021/CR500716F [80] L. N. Mander, M. M. McLachlan, J. Am. Chem. Soc. 2003, 125, 2400. [53] C. M. Williams, in Strategies and Tactics in Organic Synthesis (Ed. doi:10.1021/JA029725O M. Harmata) 2014, Vol 10, Ch. 11, pp. 249–270 (Academic Press: [81] L. N. Mander, C. M. Williams, Tetrahedron 2003, 59, 1105. New York, NY). doi:10.1016/B978-0-12-417185-5.00011-9 doi:10.1016/S0040-4020(02)01492-8 [54] J. E. Aho, P. M. Pihko, T. K. Rissa, Chem. Rev. 2005, 105, 4406. [82] A. J. Birch, J. Chem. Soc. 1944, 430. doi:10.1039/JR9440000430 doi:10.1021/CR050559N [83] L. N. Mander, in Comprehensive Organic Synthesis (Eds B. M. Trost, [55] L. F. Tietze, H. Geissler, J. A. Gewert, U. Jakobi, Synlett 1994, 511. I. Fleming) 1991, Vol 8, pp. 489–521 (Pergamon: Oxford). doi:10.1055/S-1994-22908 doi:10.1016/B978-0-08-052349-1.00237-7 [56] A. B. Smith, C. M. Adams, Acc. Chem. Res. 2004, 37, 365. [84] U. Shah, S. Chackalamannil, A. K. Ganguly, M. Chelliah, S. doi:10.1021/AR030245R Kolotuchin, A. Buevich, A. McPhail, J. Am. Chem. Soc. 2006, 128, [57] N. Choukchou-Braham, Y. Asakawa, J.-P. Lepoittevin, Tetrahedron 12654. doi:10.1021/JA065198N Lett. 1994, 35, 3949. doi:10.1016/S0040-4039(00)76710-4 [85] D. A. Evans, D. J. Adams, J. Am. Chem. Soc. 2007, 129, 1048. [58] K.-M. Chen, G. E. Hardtmann, K. Prasad, O. Repicˇ, M. J. Shapiro, doi:10.1021/JA0684996 Tetrahedron Lett. 1987, 28, 155. doi:10.1016/S0040- [86] (a) R. T. Larson, M. D. Clift, R. J. Thomson, Angew. Chem. Int. Ed. 4039(00)95673-9 2012, 51, 2481. doi:10.1002/ANIE.201108227 [59] M. Nakatsuka, J. A. Ragan, T. Sammakia, D. B. Smith, D. E. Uehling, (b) R. T. Larson, R. P. Pemberton, J. M. Franke, D. J. Tantillo, R. J. S. L. Schreiber, J. Am. Chem. Soc. 1990, 112, 5583 and references Thomson, J. Am. Chem. Soc. 2015, 137, 11197. doi:10.1021/JACS. therein. doi:10.1021/JA00170A024 5B07710 [60] L. Dong, V. A. Gordon, R. L. Grange, J. Johns, P. G. Parsons, A. [87] (a) M. D. Levin, K. Subrahamanian, H. Katz, M. B. Smith, D. J. Porzelle, P. Reddell, H. Schill, C. M. Williams, J. Am. Chem. Soc. Burlett, S. M. Hecht, J. Am. Chem. Soc. 1980, 102, 1452. doi:10.1021/ 2008, 130, 15262. doi:10.1021/JA807133P JA00524A055 [61] L. Dong, H. Schill, R. L. Grange, A. Porzelle, J. P. Johns, P. G. (b) Y. Aoyagi, K. Katano, H. Suguna, J. Primeau, L.-H. Chang, S. M. Parsons, V. A. Gordon, P. W. Reddell, C. M. Williams, Chem. – Eur. Hecht, J. Am. Chem. Soc. 1982, 104, 5537. doi:10.1021/ J. 2009, 15, 11307. doi:10.1002/CHEM.200901525 JA00384A067 [62] (a) T. M. Trnka, R. H. Grubbs, Acc. Chem. Res. 2001, 34,18. [88] T. Kametani, K. Fukumoto, Acc. Chem. Res. 1976, 9, 319. doi:10.1021/AR000114F doi:10.1021/AR50105A002 (b) S. B. Garber, J. S. Kingsbury, B. L. Gray, A. H. Hoveyda, J. Am. [89] A. Pagani, S. Gaeta, A. I. Savchenko, C. M. Williams, G. Appendino, Chem. Soc. 2000, 122, 8168. doi:10.1021/JA001179G Beilstein J. Org. Chem. 2017, 13, 1361. doi:10.3762/BJOC.13.133 [63] D. A. Evans, B. W. Trotter, P. J. Coleman, B. Coˆte´, L. C. Dias, H. A. [90] (a) Z. Liu, Z. Ding, K. Chen, M. Xu, T. Yu, G. Tong, H. Zhang, P. Li, Rajapakse, A. N. Tyler, Tetrahedron 1999, 55, 8671. doi:10.1016/ Nat. Prod. Rep. 2021, doi:10.1039/D0NP00086H S0040-4020(99)00438-X (b) S. Chow, T. Krainz, P. V. Bernhardt, C. M. Williams, Org. Lett. [64] J. R. Hanson, Sci. Prog. 2017, 100,63. doi:10.3184/ 2019, 21, 8761. doi:10.1021/ACS.ORGLETT.9B03379 003685017X14876775256129 (c) S. Chow, T. Krainz, C. J. Bettencourt, N. Broit, B. Ferguson, M. [65] R. B. Woodward, F. G. Brutschy, H. Baer, J. Am. Chem. Soc. 1948, Zhu, K. G. Hull, G. K. Pierens, P. V. Bernhardt, P. G. Parsons, D. 70, 4216. doi:10.1021/JA01192A070 Romo, G. M. Boyle, C. M. Williams, Chem. – Eur. J. 2020, 26, [66] (a) W. O. Kermack, R. Robinson, J. Chem. Soc. Trans. 1922, 121, 13372. doi:10.1002/CHEM.202003221 427. doi:10.1039/CT9222100427 [91] (a) L. N. Mander, Acc. Chem. Res. 1983, 16,48. doi:10.1021/ (b) M. Saltzman, J. Chem. Educ. 1972, 49, 750. doi:10.1021/ AR00086A003 ED049P750 (b) L. N. Mander, Nat. Prod. Rep. 1988, 5, 541. doi:10.1039/ [67] L. Birladeanu, Angew. Chem. Int. Ed. 2003, 42, 1202. doi:10.1002/ NP9880500541 ANIE.200390318 (c) L. N. Mander, Chem. Rev. 1992, 92, 573. doi:10.1021/ [68] M. Kahler, Arch. Pharm. 1830, 34, 318. doi:10.1002/ARDP. CR00012A005 18300340303 (d) L. N. Mander, Nat. Prod. Rep. 2003, 20,49. doi:10.1039/ [69] (a) G. R. Clemo, R. D. Haworth, E. J. Walton, J. Chem. Soc. 1929, B007744P 2368. doi:10.1039/JR9290002368 [92] R. D. Dawe, L. N. Mander, J. V. Turner, Tetrahedron Lett. 1985, 26, (b) G. R. Clemo, R. D. Haworth, E. J. Walton, J. Chem. Soc. 1930, 363. doi:10.1016/S0040-4039(01)80818-2 1110. doi:10.1039/JR9300001110 [93] (a) I. Ojima, I. Habus, M. Zhao, M. Zucco, Y. H. Parr, C. M. Sun, [70] J. W. Huffman, W. T. Pennington, D. W. Bearden, J. Nat. Prod. 1992, T. Brigaud, Tetrahedron 1992, 48, 6985. doi:10.1016/S0040- 55, 1087. doi:10.1021/NP50086A009 4020(01)91210-4 [71] A. P. J. Brunskill, H. W. Thompson, R. A. Lalancette, Acta Crys- (b) Y. Kanda, H. Nakamura, S. Umemiya, R. K. Puthukanoori, tallogr. C 1999, 55, 566. doi:10.1107/S0108270198014231 V. R. M. Appala, G. K. Gaddamanugu, B. R. Paraselli, P. S. Baran, The Future of Retrosynthesis 323

J. Am. Chem. Soc. 2020, 142, 10526 and references therein. [115] Y. Fukuyama, H. Minami, A. Matsuo, K. Kitamura, M. Akizuki, M. doi:10.1021/JACS.0C03592 Kubo, M. Kodama, Chem. Pharm. Bull. 2002, 50, 368 and references [94] M. Boehm, P. C. Fuenfschilling, M. Krieger, E. Kuesters, F. Struber, therein. doi:10.1248/CPB.50.368 Org. Process Res. Dev. 2007, 11, 336. doi:10.1021/OP0602425 [116] B. D. Schwartz, C. M. Williams, P. V. Bernhardt, Beilstein J. Org. [95] H. M. E. Cardwell, J. W. Cornforth, S. R. Duff, H. Holtermann, R. Chem. 2008,4. doi:10.3762/BJOC.4.34 Robinson, Chem. Ind. 1951, 389. [117] H. M. L. Davies, Ø. Loe, D. G. Stafford, Org. Lett. 2005, 7, 5561. [96] Y. Shi, J. T. Wilmot, L. U. Nordstrøm, D. S. Tan, D. Y. Gin, J. Am. doi:10.1021/OL052005C Chem. Soc. 2013, 135, 14313 and references therein. doi:10.1021/ [118] C. M. Williams, in Strategies and Tactics in Organic Synthesis (Ed. JA4064958 M. Harmata) 2012, Vol 8. Ch. 15, pp. 395–412 (Academic Press: New [97] (a) N. A. Doering, R. Sarpong, R. W. Hoffmann, Angew. Chem. Int. York, NY). doi:10.1016/B978-0-12-386540-3.00017-4 Ed. 2020, 59, 10722. doi:10.1002/ANIE.201909656 [119] B. D. Schwartz, C. M. Williams, E. Anders, P. V. Bernhardt, Tetra- (b) S. V. McCowen, N. A. Doering, R. Sarpong, Chem. Sci. 2020, 11, hedron 2008, 64, 6482. doi:10.1016/J.TET.2008.04.068 7538. doi:10.1039/D0SC01441A [120] B. D. Schwartz, J. R. Denton, P. V. Bernhardt, H. M. L. Davies, C. M. [98] For Williams group efforts in the synthesis of C20-diterpene alka- Williams, Synthesis 2009, 2840. doi:10.1055/S-0029-1216912 loids see: (a) C. M. Williams, L. N. Mander, Org. Lett. 2003, 5, 3499. [121] S. B. Jones, B. Simmons, A. Mastracchio, D. W. C. MacMillan, doi:10.1021/OL0353060 Nature 2011, 475, 183. doi:10.1038/NATURE10232 (b) C. M. Williams, L. N. Mander, P. V. Bernhardt, A. C. Willis, [122] See for example: (a) J. Wang, S.-G. Chen, B.-F. Sun, G.-Q. Lin, Y.-J. Tetrahedron 2005, 61, 3759. doi:10.1016/J.TET.2005.02.014 Shang, Chem. – Eur. J. 2013, 19, 2539. doi:10.1002/CHEM. [99] H. Zhang, M. S. Reddy, S. Phoenix, P. Deslongchamps, Angew. 201203467 Chem. Int. Ed. 2008, 47, 1272. doi:10.1002/ANIE.200704959 (b) W. Xu, J. Zhao, C. Tao, H. Wang, Y. Li, B. Cheng, H. Zhai, Org. [100] See for example: A. Greule, J. E. Stok, J. J. De Voss, M. J. Cryle, Nat. Lett. 2018, 20, 1509. doi:10.1021/ACS.ORGLETT.8B00210 Prod. Rep. 2018, 35, 757 and references therein. doi:10.1039/ (c) C. Qiao, W. Zhang, J.-C. Han, W.-M. Dai, C.-C. Li, Tetrahedron C7NP00063D 2019, 75, 1739. doi:10.1016/J.TET.2018.11.019 [101] X. Zhang, E. King-Smith, L.-Bi. Dong, L.-C. Yang, J. D. Rudolf, B. (d) D.-W. Zhang, H.-L. Fan, W. Zhang, C.-J. Li, S. Luo, H.-B. Qin, Shen, H. Renata, Science 2020, 369, 799. doi:10.1126/SCIENCE. Chem. Commun. 2020, 10066. doi:10.1039/D0CC04064A ABB8271 (e) S.-H. Wang, R.-Q. Si, Q.-B. Zhuang, X. Guo, T. Ke, X.-M. Zhang, [102] (a) D. M. Pinkerton, T. J. Vanden Berg, P. V. Bernhardt, C. M. F.-M. Zhang, Y.-Q. Tu, Angew. Chem. Int. Ed. 2020, 59, 21954. Williams, Chem. – Eur. J. 2017, 23, 2282. doi:10.1002/CHEM. doi:10.1002/ANIE.202009238 201605751 [123] (a) V. Lakshmi, P. Gupta, Nat. Prod. Res. 2008, 22, 1197. (b) D. M. Pinkerton, P. V. Bernhardt, G. P. Savage, C. M. Williams, doi:10.1080/14786410701654909 Asian J. Org. Chem. 2017, 6, 583. doi:10.1002/AJOC.201700013 (b) D. A. Mulholland, B. Parel, P. H. Coombes, Curr. Org. Chem. (c) D. M. Pinkerton, S. Chow, N. H. Eisa, K. Kainth, T. J. Vanden 2000, 4, 1011. doi:10.2174/1385272003375941 Berg, J. M. Burns, L. W. Guddat, G. P. Savage, A. Chadli, C. M. (c) J. Wu, Q. Xiao, J. Xu, M.-Y. Li, J.-Y. Pan, M.-h. Yang, Nat. Prod. Williams, Chem. – Eur. J. 2019, 25, 1451. doi:10.1002/CHEM. Rep. 2008, 25, 955. doi:10.1039/B807365A 201805420 [124] J. M. Faber, C. M. Williams, Chem. Commun. 2011, 2258. [103] IBX has sufficient acidity to promote side reactions and/or acid doi:10.1039/C0CC04698A catalysed deprotections, see ref. [28] and T. Waters, J. Boulton, T. [125] J. M. Faber, W. A. Eger, C. M. Williams, J. Org. Chem. 2012, 77, Clark, M. J. Gallen, C. M. Williams, R. A. J. O’Hair, Org. Biomol. 8913. doi:10.1021/JO301182F Chem. 2008, 6, 2530. doi:10.1039/B803452D [126] J. D’Angelo, M. B. Smith, Hybrid Retrosynthesis Organic Synthesis [104] (a) M. J. Schnermann, R. A. Shenvi, Nat. Prod. Rep. 2015, 32, 543. Using Reaxys and SciFinder 2015 (Elsevier: Amsterdam). doi:10.1039/C4NP00109E [127] G. E´ . Vle´duts, V. K. Finn, Inf. Storage Retr. 1963, 1, 101. (b) M. J. Garson, J. S. Simpson, Nat. Prod. Rep. 2004, 21, 164. doi:10.1016/0020-0271(63)90012-3 doi:10.1039/B302359C [128] M. Bersohn, A. Esack, Chem. Rev. 1976, 76, 269. doi:10.1021/ [105] E. J. Corey, P. A. Magriotis, J. Am. Chem. Soc. 1987, 109, 287. CR60300A005 doi:10.1021/JA00235A052 [129] (a) M. H. Todd, Chem. Soc. Rev. 2005, 34, 247. doi:10.1039/ [106] (a) K. A. Fairweather, L. N. Mander, Org. Lett. 2006, 8, 3395. B104620A doi:10.1021/OL061228F (b) A. Cook, A. P. Johnson, J. Law, M. Mirzazadeh, O. Ravitz, A. (b) K. A. Fairweather, S. R. Crabtree, L. N. Mander, in Strategies and Simon, WIREs Comput. Mol. Sci. 2012, 2,79.doi:10.1002/WCMS.61 Tactics in Organic Synthesis (Ed. M. Harmata) 2008, Vol 7. Ch. 2, (c) O. Engkvist, P. O. Norrby, N. Selmi, Y. Lam, Z. Peng, E. C. pp. 35–58 (Academic Press: New York, NY). doi:10.1016/S1874- Sherer, W. Amberg, T. Erhard, L. A. Smyth, Drug Discov. Today 6004(08)80006-8 2018, 23, 1203. doi:10.1016/J.DRUDIS.2018.02.014 [107] J. S. Simpson, M. J. Garson, Org. Biomol. Chem. 2004, 2, 939. (d) F. Feng, L. Lai, J. Pei, Front. Chem. 2018, 6, 199. doi:10.3389/ doi:10.1039/B315894B FCHEM.2018.00199 [108] H. Miyaoka, Y. Okubo, M. Muroi, H. Mitome, E. Kawashima, Chem. (e) F. Strieth-Kalthoff, F. Sandfort, M. H. S. Segler, F. Glorius, Chem. Lett. 2011, 40, 246. doi:10.1246/CL.2011.246 Soc. Rev. 2020, 49, 6154. doi:10.1039/C9CS00786E [109] P. C. Roosen, C. D. Vanderwal, Angew. Chem. Int. Ed. 2016, 55, (f) P. M. Pflu¨ger, F. Glorius, Angew. Chem. Int. Ed. 2020, 59, 18860. 7180. doi:10.1002/ANIE.201603581 doi:10.1002/ANIE.202008366 [110] E. E. Robinson, R. J. Thomson, J. Am. Chem. Soc. 2018, 140, 1956. (g) W.-D. Ihlenfeldt, J. Gasteiger, Angew. Chem. Int. Ed. Engl. 1996, doi:10.1021/JACS.7B13234 34, 2613. doi:10.1002/ANIE.199526131 [111] A. S. Karns, B. D. Ellis, P. C. Roosen, Z. Chahine, K. G. Le Roch, (h) G. Schneider, D. E. Clark, Angew. Chem. Int. Ed. 2019, 58, 10792. C. D. Vanderwal, Angew. Chem. Int. Ed. 2019, 58, 13749. doi:10.1002/ANIE.201814681 doi:10.1002/ANIE.201906834 (i) A. Thakkar, S. Johansson, K. Jorner, D. Buttar, J.-L. Reymond, O. [112] The Claisen Rearrangement: Methods and Applications (Eds M. Engkvist, React. Chem. Eng. 2021, 6,27. doi:10.1039/D0RE00340A Hiersemann, U. Nubbemeyer) 2007 (Wiley-VCH: Chichester). [130] (a) A. Bondy, U. S. R. Murty, Graph Theory 2008 (Springer-Verlag: [113] B. D. Schwartz, J. R. Denton, H. M. L. Davies, C. M. Williams, Aust. London). J. Chem. 2009, 62, 980. doi:10.1071/CH09267 (b) L. David, A. Thakkar, R. Mercado, O. Engkvist, J. Cheminform. [114] B. D. Schwartz, J. R. Denton, Y. Lian, H. M. L. Davies, C. M. 2020, 12,56. doi:10.1186/S13321-020-00460-5 Williams, J. Am. Chem. Soc. 2009, 131, 8329. doi:10.1021/ [131] T. Fink, H. Bruggesser, J.-L. Reymond, Angew. Chem. Int. Ed. 2005, JA9019484 44, 1504. doi:10.1002/ANIE.200462457 324 C. M. Williams and M. A. Dallaston

[132] IBM Corporation, IBM: Icons of Progress. Deep Blue. Available [152] (a) B. A. Grzybowski, S. Szymkuc´, E. P. Gajewska, K. Molga, P. at: https://www.ibm.com/ibm/history/ibm100/us/en/icons/deepblue Dittwald, A. Wołos, T. Klucznik, Chem 2018, 4, 390. doi:10.1016/ (accessed 18 Nov 2020) J.CHEMPR.2018.02.024 [133] (a) D. Silver, J. Schrittwieser, K. Simonyan, I. Antonoglou, A. (b) E. P. Gajewska, S. Szymkuc´, P. Dittwald, M. Startek, O. Popik, J. Huang,A.Guez,T.Hubert,L.Baker,M.Lai,A.Bolton,Y. Mlynarski, B. A. Grzybowski, Chem 2020, 6, 280. doi:10.1016/ Chen,T.Lillicrap,F.Hui,L.Sifre,G.vandenDriessche,T. J.CHEMPR.2019.11.016 Graepel, D. Hassabis, Nature 2017, 550,354.doi:10.1038/ [153] C. M. Gothard, N. A. Gothard, B. A. Grzybowski, in Abstracts of NATURE24270 Papers, 245th ACS National Meeting & Exposition (American (b) D. Silver, T. Hubert, J. Schrittwieser, I. Antonoglou, M. Lai, A. Chemical Society: New Orleans, LA) 2013,COMP276.Poster Guez, M. Lanctot, L. Sifre, D. Kumaran, T. Graepel, T. Lillicrap, K. abstract. Available at https://tpa.acs.org/abstract/245tNM-353448/ Simonyan, D. Hassabis, Science 2018, 362, 1140. doi:10.1126/SCI analysis-of-chemical-networks-a-novel-paradigm-for-optimal- ENCE.AAR6404 synthesis (c) M. MacDonald, How Computers Are Reinventing Chess. 15 Nov [154] B. Liu, B. Ramsundar, P. Kawthekar, J. Shi, J. Gomes, Q. Luu 2019. Available at: https://medium.com/young-coder/how-compu- Nguyen, S. Ho, J. Sloane, P. Wender, V. Pande, ACS Cent. Sci. 2017, ters-play-chess-today-d4f921ce8c4c (accessed 10 Feb 2021) 3, 1103. doi:10.1021/ACSCENTSCI.7B00303 [134] D. Heaven, Nature 2019, 574,163. doi:10.1038/D41586-019- [155] (a) P. Schwaller, T. Gaudin, D. La´nyi, C. Bekas, T. Laino, Chem. Sci. 03013-5 2018, 9, 6091. doi:10.1039/C8SC02339E [135] E. J. Corey, Q. Rev. Chem. Soc. 1971, 25, 455. doi:10.1039/ (b) A. C. Vaucher, F. Zipoli, J. Geluykens, V. H. Nair, P. Schwaller, QR9712500455 T. Laino, Nat. Commun. 2020, 11, 3601. doi:10.1038/S41467-020- [136] (a) T. Klucznik, B. Mikulak-Klucznik, M. P. McCormack, H. Lima, 17266-6 S. Szymkuc´, M. Bhowmick, K. Molga, Y. Zhou, L. Rickershauser, [156] IBM Corporation, ReactionSeq2Seq_Dataset. Available at: https:// E. P. Gajewska, A. Toutchkine, P. Dittwald, M. P. Startek, G. J. ibm.box.com/v/ReactionSeq2SeqDataset (accessed 4 Dec 2020) Kirkovits, R. Roszak, A. Adamski, B. Sieredzin´ska, M. Mrksich, [157] I. A. Watson, J. Wang, C. A. Nicolaou, J. Cheminform. 2019, 11,1. S. L. J. Trice, B. A. Grzybowski, Chem 2018, 4, 522. doi:10.1016/J. doi:10.1186/S13321-018-0323-6 CHEMPR.2018.02.002 [158] Spaya, a community-enriched algorithm for data-driven retrosyn- (b) K. Molga, S. Szymkuc´, B. A. Grzybowski, Acc. Chem. Res. 2021, thetic planning developed by Iktos, spaya.ai, 2020. 54, 1094. doi:10.1021/ACS.ACCOUNTS.0C00714 [159] S. Genheden, A. Thakkar, V. Chadimova´, J.-L. Reymond, O. [137] J. B. Hendrickson, Anal. Chim. Acta 1990, 235, 103. doi:10.1016/ Engkvist, E. Bjerrum, J. Cheminform. 2020, 12,70. doi:10.1186/ S0003-2670(00)82066-8 S13321-020-00472-1 [138] (a) J. Gasteiger, C. Jochum, in Organic Compounds. Topics in [160] E. J. Corey, M. Ohno, R. B. Mitra, P. A. Vatakencherry, J. Am. Chem. Current Chemistry 1978, Vol 74, pp. 93–126 (Springer: Berlin). Soc. 1964, 86, 478. doi:10.1021/JA01057A039 doi:10.1007/BFB0050147 [161] W. L. Jorgensen, E. R. Laird, A. J. Gushurst, J. M. Fleischer, S. A. (b) J. Gasteiger, M. G. Hutchings, B. Christoph, L. Gann, C. Hiller, P. Gothe, H. E. Helson, G. D. Paderes, S. Sinclair, Pure Appl. Chem. Lo¨w, M. Marsili, H. Saller, K. Yuki, in Organic Synthesis, Reactions 1990, 62, 1921. doi:10.1351/PAC199062101921 and Mechanisms. Topics in Current Chemistry 1987, Vol 137, [162] J. M. Fleischer, A. J. Gushurst, W. L. Jorgensen, J. Org. Chem. 1995, pp. 19–73 (Springer: Berlin). doi:10.1007/3-540-16904-0_14 60, 490. doi:10.1021/JO00108A006 [139] T. D. Salatin, W. L. Jorgensen, J. Org. Chem. 1980, 45, 2043. [163] W. L. Jorgensen, J. Phys. Chem. B 2015, 119, 624. doi:10.1021/ doi:10.1021/JO01299A001 JP510442J [140] (a) S. Hanessian, J. Franco, B. Larouche, Pure Appl. Chem. 1990, 62, [164] M. Kubat, An Introduction to Machine Learning, 2nd Edn 2017 1887. doi:10.1351/PAC199062101887 (Springer International Publishing: Cham, Switzerland). doi:10.1007/ (b) S. Hanessian, J. Org. Chem. 2012, 77, 6657. doi:10.1021/ 978-3-319-63913-0 JO300902M [165] S. Lemonick, ‘CAS opens data vault to MIT scientists’, C&EN [141] W. T. Wipke, D. Rogers, J. Chem. Inf. Comput. Sci. 1984, 24,71. Chemical & Engineering News, 10 November 2020. Available at: doi:10.1021/CI00042A005 https://cen.acs.org/physical-chemistry/computational-chemistry/ [142] (a) J. B. Hendrickson, A. G. Toczko, J. Chem. Inf. Comput. Sci. 1989, CAS-opens-data-vault-MIT/98/web/2020/11. doi:10.47287/CEN-09844- 29, 137. doi:10.1021/CI00063A001 SCICON1 (b) J. B. Hendrickson, A. G. Toczko, Pure Appl. Chem. 1989, 61, 589. [166] A. Thakkar, T. Kogej, J.-L. L. Reymond, O. Engkvist, E. J. Bjerrum, doi:10.1351/PAC198961030589 Chem. Sci. 2020, 11, 154. doi:10.1039/C9SC04944D [143] E. S. Blurock, J. Chem. Inf. Comput. Sci. 1990, 30, 505. doi:10.1021/ [167] E. J. Corey, S. Nozoe, J. Am. Chem. Soc. 1965, 87, 5728. doi:10.1021/ CI00068A024 JA00952A035 [144] J. Gasteiger, W. D. Ihlenfeldt, in Software Development in Chemistry [168] For other examples of direct protection group interconversion, see: 4 (Ed. J. Gasteiger) 1990, pp. 57–65 (Springer: Berlin). doi:10.1007/ C. M. Williams, L. N. Mander, Tetrahedron Lett. 2004, 45, 667. 978-3-642-75430-2_7 doi:10.1016/J.TETLET.2003.11.065 [145] G. Sello, J. Chem. Inf. Model. 1992, 32, 713. doi:10.1021/ [169] E. J. Corey, S. G. Pyne, W.-g. Su, Tetrahedron Lett. 1983, 24, 4883. CI00010A019 doi:10.1016/S0040-4039(01)99801-6 [146] H. Satoh, K. Funatsu, J. Chem. Inf. Comput. Sci. 1995, 35,34. [170] S. Hanessian, Organic Chemistry Series, Volume 3: Total Synthesis of doi:10.1021/CI00023A005 Natural Products: The ‘Chiron’ Approach 1983 (Pergamon Press: [147] G. Mehta, R. Barone, M. Chanon, Eur. J. Org. Chem. 1998, 1409. New York, NY). doi:10.1002/(SICI)1099-0690(199807)1998:7%3C1409::AID- [171] (a) M. H. S. Segler, M. Preuss, M. P. Waller, Nature 2018, 555, 604. EJOC1409%3E3.0.CO;2-H doi:10.1038/NATURE25978 [148] K. Satoh, K. Funatsu, J. Chem. Inf. Comput. Sci. 1999, 39, 316. (b) J. N. Wei, D. Duvenaud, A. Aspuru-Guzik, ACS Cent. Sci. 2016, doi:10.1021/CI980147Y 2, 725. doi:10.1021/ACSCENTSCI.6B00219 [149] I. M. Socorro, K. Taylor, J. M. Goodman, Org. Lett. 2005, 7, 3541. (c) H. Weir, K. Thompson, B. Choi, A. Woodward, A. Braun, doi:10.1021/OL0512738 T. J. Martı´nez, ChemRxiv 2021. doi:10.26434/CHEMRXIV. [150] A. Bøgevig, H.-J. Federsel, F. Huerta, M. G. Hutchings, H. Kraut, T. 14156957.V1 Langer, P. Lo¨w, C. Oppawsky, T. Rein, H. Saller, Org. Process Res. [172] (a) C. W. Coley, L. Rogers, W. H. Green, K. F. Jensen, ACS Cent. Sci. Dev. 2015, 19, 357. doi:10.1021/OP500373E 2017, 3, 1237. doi:10.1021/ACSCENTSCI.7B00355 [151] M. A. Kayala, P. Baldi, J. Chem. Inf. Model. 2012, 52, 2526. (b) C. W. Coley, W. H. Green, K. F. Jensen, Acc. Chem. Res. 2018, 51, doi:10.1021/CI3003039 1281. doi:10.1021/ACS.ACCOUNTS.8B00087 The Future of Retrosynthesis 325

(c) H. Gao, T. J. Struble, C. W. Coley, Y. Wang, W. H. Green, K. F. [187] A. M. Dilmaç, E. Spuling, A. de Meijere, S. Bra¨se, Angew. Chem. Int. Jensen, ACS Cent. Sci. 2018, 4, 1465. doi:10.1021/ACSCENTSCI. Ed. 2017, 56, 5684. doi:10.1002/ANIE.201603951 8B00357 [188] (a) M. A. Sierra, M. C. de la Torre, Angew. Chem. Int. Ed. 2000, 39, (d) C. W. Coley, T. J. Struble, H. Gao, X. Wang, W. Lin, R. Barzilay, 1538. doi:10.1002/(SICI)1521-3773(20000502)39:9,1538::AID- T. Jaakkola, W. H. Green, K. F. Jensen, in Abstracts of Papers, 258th ANIE1538.3.0.CO;2-O ACS National Meeting & Exposition (American Chemical Society: (b) M. A. Sierra, M. C. de la Torre, Dead Ends and Detours: Direct San Diego, CA, USA) 2019, ORGN 274. Poster abstract. Available Ways to Successful Total Synthesis 2004 (Wiley-VCH: Weinheim). at: https://tpa.acs.org/abstract/acsnm258-3212520/machine-learn- [189] (a) M. Wernerova, T. Hudlicky´, Synlett 2010, 2701. doi:10.1055/S- ing-for-organic-chemistry-reaction-prediction-and-retrosynthesis 0030-1259018 (e) M. E. Fortunato, C. W. Coley, B. C. Barnes, K. F. Jensen, J. Chem. (b) T. Hudlicky´, Chem. Rev. 1996, 96,3. doi:10.1021/CR950012G Inf. Model. 2020, 60, 3398. doi:10.1021/ACS.JCIM.0C00403 [190] (a) P. J. Kocien´ski, Protecting Groups, 3rd Edn 2005 (Georg Thieme [173] (a) K. C. Nicolaou, S. A. Snyder, Angew. Chem. Int. Ed. 2005, 44, Verlag: Stuttgart). 1012. doi:10.1002/ANIE.200460864 (b) For another leading text in this regard see: Greene’s Protective (b) M. E. Maier, Nat. Prod. Rep. 2009, 26, 1105. doi:10.1039/ Groups in Organic Synthesis, 5th Edn (Ed. P. G. M. Wuts) 2014 (John B809658A Wiley & Sons Inc.: Hoboken, NJ). doi:10.1002/9781118905074 (c) Y. Usami, Mar. Drugs 2009, 7, 314. doi:10.3390/MD7030314 [191] B. Mikulak-Klucznik, P. Gołebiowska,˛ A. A. Bayly, O. Popik, T. (d) T. L. Suyama, W. H. Gerwick, K. L. McPhail, Bioorg. Med. Chem. Klucznik, S. Szymkuc´, E. P. Gajewska, P. Dittwald, O. Staszewska- 2011, 19, 6675. doi:10.1016/J.BMC.2011.06.011 Krajewska, W. Beker, T. Badowski, K. A. Scheidt, K. Molga, J. (e) B. Khatri Chhetri, S. Lavoie, A. M. Sweeney-Jones, J. Kubanek, Mlynarski, M. Mrksich, B. A. Grzybowski, Nature 2020, 588,83. Nat. Prod. Rep. 2018, 35, 514. doi:10.1039/C8NP00011E doi:10.1038/S41586-020-2855-Y [174] (a) P. Wipf, Y. Uto, J. Org. Chem. 2000, 65, 1037. doi:10.1021/ [192] (a) I. S. Young, P. S. Baran, Nat. Chem. 2009, 1, 193. doi:10.1038/ JO9914566 NCHEM.216 (b) P. Wipf, Y. Uto, Tetrahedron Lett. 1999, 40, 5165. doi:10.1016/ (b) For an example of a protecting group free synthesis of an S0040-4039(99)00957-0 advanced intermediate, see: R. Heim, S. Wiedemann, C. M. (c) B. McKeever, G. Pattenden, Tetrahedron Lett. 2001, 42, 2573. Williams, P. V. Bernhardt, Org. Lett. 2005, 7, 1327. doi:10.1021/ doi:10.1016/S0040-4039(01)00194-0 OL0501222 (d) B. McKeever, G. Pattenden, Tetrahedron 2003, 59, 2713. [193] J. Mulzer, Nat. Prod. Rep. 2014, 31, 595. doi:10.1039/C3NP70105K doi:10.1016/S0040-4020(03)00294-1 [194] See for example: C. D. G. Read, P. W. Moore, C. M. Williams, Green [175] B. Verbraeken, J. Hullaert, J. van Guyse, K. Van Hecke, J. Winne, R. Chem. 2015, 17, 4537. doi:10.1039/C5GC01346A Hoogenboom, J. Am. Chem. Soc. 2018, 140, 17404. doi:10.1021/ [195] C&EN Media Group, SYNTHIATM Retrosynthesis Software for JACS.8B10918 Practicing Chemists: A Look at Recent Applications in the Labora- [176] Rita, M. H. bin Abdul Rahman, S. S. M. Chong, R. W. Bates, Synlett tory 15 Sep 2020. Available at: https://cen.acs.org/media/webinar/ 2020, 31, 1479. doi:10.1055/S-0040-1707470 milliporesigma_091520.html (accessed 28 Sep 2020) [177] L. A. Maslovskaya, A. I. Savchenko, E. H. Krenske, S. Chow, T. Holt, [196] R. Chen, Y. Shen, S. Yang, Y. Zhang, Angew. Chem. Int. Ed. 2020, V. A. Gordon, P. W. Reddell, C. J. Pierce, P. G. Parsons, G. M. Boyle, 59, 14198. doi:10.1002/ANIE.202003735 A. G. Kutateladze, C. M. Williams, Chem. – Eur. J. 2020, 26, 11862. [197] I. Fleming, D. Williams, Spectroscopic Methods in Organic Chemis- doi:10.1002/CHEM.202001884 try, 7th Edn 2019 (Springer: Cham, Switzerland). doi:10.1007/978-3- [178] D. Zhang, D. Ruan, J. Li, Z. Chen, W. Zhu, F. Guo, K. Chen, Y. Li, R. 030-18252-6 Wang, Phytochemistry 2020, 174, 112337. doi:10.1016/J.PHY [198] P. J. Stevenson, Org. Biomol. Chem. 2011, 9, 2078. doi:10.1039/ TOCHEM.2020.112337 C0OB00705F [179] A. I. Savchenko, C. M. Williams, Eur. J. Org. Chem. 2013, 7263. [199] T. J. Vanden Berg, D. M. Pinkerton, C. M. Williams, Org. Biomol. doi:10.1002/EJOC.201301308 Chem. 2017, 15, 7102. doi:10.1039/C7OB01400G [180] (a) J. Y. W. Mak, R. H. Pouwer, C. M. Williams, Angew. Chem. Int. [200] (a) K. M. Peese, D. Y. Gin, J. Am. Chem. Soc. 2006, 128, 8734. Ed. 2014, 53, 13664. doi:10.1002/ANIE.201400932 doi:10.1021/JA0625430 (b) E. H. Krenske, C. M. Williams, Angew. Chem. Int. Ed. 2015, 54, (b) K. M. Peese, D. Y. Gin, Chem. – Eur. J. 2008, 14, 1654. 10608. doi:10.1002/ANIE.201503822 doi:10.1002/CHEM.200701290 (c) A. G. Kutateladze, E. H. Krenske, C. M. Williams, Angew. Chem. (c) The Williams group has taken a second-order approach to this Int. Ed. 2019, 58, 7107. doi:10.1002/ANIE.201902777 system.[98] [181] M. Jaspars, Nat. Prod. Rep. 1999, 16, 241. doi:10.1039/A804433C [201] J. Li, M. D. Eastgate, Org. Biomol. Chem. 2015, 13, 7164. [182] S. G. Smith, J. M. Goodman, J. Am. Chem. Soc. 2010, 132, 12946. doi:10.1039/C5OB00709G doi:10.1021/JA105035R [202] (a) T. J. Zerk, P. W. Moore, C. M. Williams, P. V. Bernhardt, Chem. [183] (a) A. G. Kutateladze, D. M. Kuznetsov, A. A. Beloglazkina, T. Holt, Commun. 2016, 10301. doi:10.1039/C6CC05440D J. Org. Chem. 2018, 83, 8341. doi:10.1021/ACS.JOC.8B01027 (b) T. J. Zerk, P. W. Moore, J. S. Harbort, S. Chow, L. Byrne, G. A. (b) For additional DU8þ structural revisions see: T. A. Holt, D. S. Koutsantonis, J. R. Harmer, M. Martıńez, C. M. Williams, P. V. Reddy, D. B. Huple, L. M. West, A. D. Rodriguez, M. T. Crimmins, Bernhardt, Chem. Sci. 2017, 8, 8435. doi:10.1039/C7SC04260D A. G. Kutateladze, J. Org. Chem. 2020, 85, 6201 and references (c) P. W. Moore, T. J. Zerk, J. M. Burns, P. V. Bernhardt, C. M. therein. doi:10.1021/ACS.JOC.0C00555 Williams, Eur. J. Org. Chem. 2019, 303. doi:10.1002/EJOC. [184] (a) M. M. Zanardi, F. A. Biglione, M. A. Sortino, A. M. Sarotti, 201800860 J. Org. Chem. 2018, 83, 11839. doi:10.1021/ACS.JOC.8B01749 [203] (a) R. H. Grubbs, S. J. Miller, G. C. Fu, Acc. Chem. Res. 1995, 28, 446. (b) N. Grimblat, J. A. Gavıń, A. H. Daranas, A. M. Sarotti, Org. Lett. doi:10.1021/AR00059A002 2019, 21, 4003. doi:10.1021/ACS.ORGLETT.9B01193 (b) O. M. Ogba, N. C. Warner, D. J. O’Leary, R. H. Grubbs, Chem. [185] M. O. Marcarino, M. M. Zanardi, A. M. Sarotti, Org. Lett. 2020, 22, Soc. Rev. 2018, 47, 4510. doi:10.1039/C8CS00027A 3561. doi:10.1021/ACS.ORGLETT.0C01001 [204] (a) B. Burger, P. M. Maffettone, V. V. Gusev, C. M. Aitchison, Y. [186] (a) J. M. Humphrey, Y. Liao, A. Ali, T. Rein, Y.-L. Wong, H.-J. Chen, Bai, X. Wang, X. Li, B. M. Alston, B. Li, R. Clowes, N. Rankin, B. A. K. Courtney, S. F. Martin, J. Am. Chem. Soc. 2002, 124, 8584 and Harris, R. S. Sprick, A. I. Cooper, Nature 2020, 583, 237. references therein. doi:10.1021/JA0202964 doi:10.1038/S41586-020-2442-2 (b) S. F. Martin, H.-J. Chen, A. K. Courtney, Y. Liao, M. Pa¨tzel, M. N. (b) C. W. Coley, N. S. Eyke, K. F. Jensen, Angew. Chem. Int. Ed. Ramser, A. S. Wagman, Tetrahedron 1996, 52, 7251. doi:10.1016/ 2020, 59, 22858. doi:10.1002/ANIE.201909987 0040-4020(96)00249-9 [205] A. Turing, Mind 1950, LIX, 433. doi:10.1093/MIND/LIX.236.433 326 C. M. Williams and M. A. Dallaston

[206] J. R. Searle, Behav. Brain Sci. 1980, 3, 417. doi:10.1017/ [213] P. Judson, Knowledge-based Expert Systems in Chemistry: Not S0140525X00005756 Counting on Computers 2009 (RSC: Cambridge). doi:10.1039/ [207] J. R. Searle, Proc. Addresses Am. Philos. Assoc. 1990, 64,21. 9781847559807 doi:10.2307/3130074 [214] H. Else, ‘Need to make a molecule? Ask this AI for [208] C. M. Signorelli, Front. Robot. AI 2018, 5, 121. doi:10.3389/FROBT. instructions’, Nature News, 28 March 2018. doi:10.1038/D41586- 2018.00121 018-03977-W [209] E. A. Feigenbaum, J. ACM 2003, 50,32. doi:10.1145/602382.602400 [215] A. R. Battersby, D. W. Young, Biogr. Mem. Fellows R. Soc. 2016, 62, [210] See for example: P. A. Wender, B. L. Miller, Nature 2009, 460, 197. 19. doi:10.1098/RSBM.2015.0016 doi:10.1038/460197A [211] J. W. Cornforth, Aust. J. Chem. 1993, 46, 157. doi:10.1071/ CH9930157 [212] D. Seebach, Angew. Chem. Int. Ed. Engl. 1990, 29, 1320. doi:10.1002/ANIE.199013201 Handling Editor: Curt Wentrup