Challenges in Creating Online Exercises and Exams in Organic Chemistry

Home , ChemDraw, Structural formula, Tropylium cation

48 CHIMIA 2018, 72, No. 1/2 ChemiCal eduCation

doi:10.2533/chimia.2018.48 Chimia 72 (2018) 48–54 © Swiss Chemical Society Challenges in Creating Online Exercises and Exams in Organic Chemistry

Bernhard Jaun and Carlo Thilgen*

Abstract: e-Learning has become increasingly important in chemical education and online exams can be an at- tractive alternative to traditional exams written on paper, particularly in classes with a large number of students. Ten years ago, we began to set up an e-course complementing our lecture courses Organic Chemistry I and II within the open-source e-learning environment Moodle. In this article, we retrace a number of decisions we took over time, thereby illustrating the challenges one faces when creating online exercises and exams in (organic) chemistry. Special emphasis is put on the development of MOSFECCS (MOlecular Structural Formula Editor and Calculator of Canonical SMILES), our new editor for drawing structural formulae and converting them to alphanumeric SMILES codes that can be submitted as answers to e-problems. Convinced that the possibility for structure input is essential to set up sensible chemistry quizzes and exams, and realising that existing tools present major flaws in an educational context, we decided to embark on the implementation of MOSFECCS which takes into account a number of didactic aspects. Keywords: e-Learning · Moodle · Online exercises and exams · SMILES generator · Structural formula editor

and its application to problems relevant fullerenes and fullerene derivatives. in biology. He is a recipient of the Silver More recently, his interest has focused Medal of ETH Zurich and the Werner on the generation of nanostructured Prize of the Swiss Chemical Society. surfaces by assembly of porous molecular He taught courses in 1st year organic networks, the ordering of which is based chemistry (1991–2012) for students of on supramolecular synthons. Carlo Biology, Pharmaceutical Sciences and Thilgen taught an advanced course in Health Sciences and Technology, as well Supramolecular Chemistry next to basic as advanced NMR and advanced physical Organic Chemistry courses for students in organicchemistryforstudentsinChemistry Chemistry and Interdisciplinary Natural and Interdisciplanary Natural Sciences. Sciences. Currently, he is teaching first- Bernhard Jaun is honorary professor year Organic Chemistry for students in (emeritus) in the Laboratory of Organic Biology, Pharmaceutical Sciences and Chemistry, ETH Zürich. After studies Health Sciences and Technology. In in chemistry at ETH Lausanne and ETH addition, he directs basic and advanced Zurich, Bernhard Jaun obtained his practical Organic Chemistry courses for PhD from ETH Zurich working under students in Biology and Pharmaceutical the supervision of A. Eschenmoser and Sciences. J.F.M Oth. After a two-year stay as a postdoctoral fellow in R. Breslow’s group at Columbia University, New York, he 1. Strengthening a Basic Organic returned to ETH Zurich where he started Chemistry Course by Online his own research group in 1989. He was Exercises: Initiation of Our Project head of the NMR-facility of the laboratory for organic chemistry from 1983 to 2013. In a basic one-year organic chemistry Bernhard Jaun’s research focus was in Carlo Thilgen received his diploma course at the university, the students are the area of physical organic chemistry degree in Chemistry and his PhD degree expected to: from the Rheinische Friedrich-Wilhelms- • learn a new ‘language’ (proper termi- Universität Bonn, working under the nology) and a new ‘script’ (structural supervision of Fritz Vögtle. After a formulae) to describe the structure and postdoctoral stay in the group of François behaviour of molecules; Diederich at the University of California, • train their perception of 3D molecular Los Angeles, he moved with this group structure and deduce it from 2D struc- to ETH Zurich, where he became Senior tural formulae; interconvert various 2D Assistant, Lecturer and, in 2011, Adjunct representations of 3D structures (per- *Correspondence: Prof. Dr. C. Thilgen Professor (Titularprofessor). He carried spective drawings, Natta-, sawhorse-, Laboratorium für Organische Chemie out research in the fields of molecular Newman, Fischer projections); ETH Zurich recognition and fullerenes, in particular the • apply rules and conventions, e.g. sys- Vladimir-Prelog-Weg 3 CH-8093 Zurich isolation and the chemistry of new higher tematic nomenclature, stereodescrip- E-mail: [email protected] fullerenes as well as the chirality of pristine tors, or Fischer projections; ChemiCal eduCation CHIMIA 2018, 72, No. 1/2 49

• carry out a qualitative conformational any time. Last but not least, we had the idea 2. A Major Flaw of Most e-Learn- analysis of basic systems, e.g. cyclo- at the back of our mind to eventually ex- ing Environments: The Missing hexane; tend the project toward the elaboration of Possibility for Molecular Structure • describe the electronic delocalization online exams. With about 500 candidates Input in π-systems by use of resonance struc- per year, computer exams appear as an ap- tures; pealing alternative to the traditional manu- As nicely illustrated by Fig. 1, chemis- • recognize typical structural sub-units ally corrected exams written on paper. try – and organic chemistry in particular – (functional groups) which confer dis- The project was initiated in the fall is a very ‘visual’ science. To convey the tinctive physical or chemical (acidity, semester 2007 by one of us (B. J.) with a essential information about a molecular reactivity) properties to molecules; relatively small set of online exercises em- structure, chemists use structural formu- • familiarize themselves with a number bedded within the e-learning environment lae, i.e. two-dimensional drawings that are of basic organic reactions and apply Moodle. In 2008, a FILEP grant (‘Fonds combinations of graphs, letters, numbers them to synthetic problems. zur Finanzierung lehrbezogener Projekte’) and other symbols, possibly including de- Whereas the necessary concepts and from the rector of the ETH Zurich allowed tails on the 3D-arrangement of atoms (ste- theoretical background to achieve these a nine-month dedicated full-time employ- reo-information). goals are usually imparted in a lecture (2–4 ment of a chemist, thereby considerably in- Because structural formulae and re- h per week) or can be learnt from textbooks, creasing the number and quality of the quiz- action schemes are so essential in organic the acquirement and mastering of the men- zes (for an example, see Fig. 5). In 2012, the chemistry, students must practice this no- tioned skills is a matter of practice. At the class was taken over by C. T. who started to tation or ‘script’ from the very beginning. Department of Chemistry and Applied build a pool of synthesis problems for the Any reasonable system for online exercises Biosciences of the ETH Zurich, this is tra- 2nd semester, in which fragmentary reaction must, therefore, accept structural formulae ditionally done in a weekly problem-solv- schemes have to be completed with prod- in the formulation of questions as well as the ing session (1–2 h), in which small groups ucts, reactants or reaction conditions. At the input of answers. Regarding the former re- of students coached by a more advanced same time, we started to convert old ‘paper quirement, the ‘quiz’ modules of most cur- fellow discuss the solutions of assignments exams’ to online problems in order to test rent e-learning environments allow the use distributed beforehand as homework. the suitability of various types of questions of graphic elements as part of the question. However, increasing student numbers in view of an online assessment. With the By contrast, it is a real challenge to provide and the policy to group students from gained experience and free infrastructure a tool that allows students to draw struc- different study programs for introductory capacities for a large online exam having tural formulae as answers and have these courses in basic science, taught as subsidi- finally become available, we are now all set checked by the computer for correctness. In ary subjects, has led to huge classes of up to for the Winter 2018 session. our view, such a graphical input tool is key 500 students. This has made it increasingly The first version of the open-source for didactic reasons, as it makes a huge dif- difficult to maintain a culture of discussion e-learning management software Moodle ference as to whether students actively try and dialogue in the exercise groups where was released in 2002[1,2] and a distribution to draw the correct structure themselves or attendance has reached a level of about 25 has been available at the ETH Zurich for simply choose one out of a given set (multi- students. about ten years. Accordingly, many people ple choice). Unfortunately, most e-learning Furthermore, we feel that first-year are familiar with the system and the pos- environments allow only multiple-choice students, who come from different schools sibilities it offers. Rather than providing type answers or the input of text or numbers with different core subjects, who have cho- a tour d’horizon of our course, we there- (alphanumeric strings) which are easily and sen different major subjects at the univer- fore decided to focus on our strategy to reliably checked for correctness. sity and head toward a variety of special- overcome a very serious limitation of the In contrast to human beings, however, izations, have extremely different back- text-based Moodle system when it comes software is quite inept at deciding whether grounds and interests and, consequently, to chemistry, i.e. the missing possibility for two drawings correspond to the same struc- need different extents of training to be a graphical input of molecular structure, an ture or not.[3] The conventions for struc- enabled to pass a uniform exam at the end essential aspect in the context of organic tural formulae are rather loose and allow of the first year. Besides, we noticed that chemistry exercises. certain things, considered as redundant, to there is a need for additional practice out- be left out, e.g. hydrogen atoms at carbon side of the lecture period, in particular during the summer break when the students prepare for the exam. These considerations – but also the curiosity about the feasibility of such an approach – prompted us to develop a se- ries of online trainings in the context of the Organic Chemistry I and II courses (1st and 2nd semester) for the students of biology, pharmaceutical sciences, and health sciences and technology. They were meant as supplementary training, not as a substi- tute for the regular assignments discussed in the weekly problem-solving session. The ludic, quiz-type character that online problems can be given in an appropriate e-learning environment, in combination with the recent mushrooming dissemina- Fig. 1. Chemists communicate through structural formulae. Photograph of the blackboard after tion of tablet computers, makes them at- a lecture of R. B. Woodward on the total synthesis of Cephalosporin C at ETH Zurich (ca. 1966; tractive brain-teasers available anywhere, Photo: courtesy of Prof. D. Arigoni). 50 CHIMIA 2018, 72, No. 1/2 ChemiCal eduCation

or even at heteroatoms, or to replace the graph of substructures by abbreviations such as R1, Ph, or by a partial molecular Structural Formula Editor formula such as NO , CO H, or COOH. 2 2 Non-bonding electron pairs may be drawn Interac(ve drawing Interac(ve drawing tool (Lewis structures), but are normally not. tool incorporated into In other respects the conventions are very ques(on form Biunique strict: omitting a formal charge on an atom encoding (e.g. H C–O instead of H C–O–) or draw- 3 3 ing a filled wedge instead of a dashed one Encoding in Alphanumeric string is a fatal error and no longer represents the Submit proprietary (SMILES,InChI etc.) same molecule. format

3. Initial Situation and First Phase Copy/paste of Our e-Learning Project Analysis and feedback

An e-learning system programmed Submit to quiz-module for online exercises in chemistry ideal- DedicatedChemistry Analysis and feedback ly includes a tool for drawing structural e-Learning Environment formulae or whole reaction schemes ei- Standard e-Learning (e.g.ACEorganic) ther directly in the question form or in a Environment (e.g. Moodle) separate window. When the drawing is submitted, the corresponding structural information is sent to the server in a Fig. 2. Two concepts for online exercises in chemistry. Left: integrated e-learning package speci- fically programmed for chemistry. Right: Modular concept using the quiz module of a standard suitable format. Server-side software then e-learning package in combination with pre-existing tools for drawing and alphanumeric encoding. analyses the answer and returns a feedback (Fig. 2, left). When we started to think about creating a pool of online quizzes, EPOCH, an 3.1 E-learning Environment ming interface). Because JME appeared e-learning package for organic chemistry We opted for Moodle, one of the slightly more stable in our hands, and P. Ertl coming close to this ideal, had just been e-learning environments that was run and was kind enough to place the package at created by R. B. Grossman and R. Finkel[4] maintained at the time (2006/7) by the our disposal, we opted for JME, which was in collaboration with ChemAxon, the devel- teaching support section of the IT services installed outside the Moodle server space opers of Marvin.[5] We had initially hoped (now LET, Educational Development and but could be called from within Moodle. to be able to use EPOCH or at least the Technology) at ETH Zurich. Our choice JME had a built-in SMILES-generator (see same standards/interface and tools for our was motivated by the fact that Moodle was below), displayed the SMILES code of the course on the ETH server. Unfortunately, open source and offered a simple but robust drawn molecule(s) in a text window and the authors sold the rights to a large pub- interface already at these early stages. We allowed its transfer to the answer field of lishing company, who renamed the pack- hoped that a system used and maintained Moodle quizzes via copy/paste. This com- age ACEorganic and offered it to the buy- by a broad open community had a better bination worked well until a few years ago, ers of its text books until 2015. As of late, chance of surviving in the rapidly chang- when increasing concerns about the securi- ACEorganic by Grossman and Finkel is ing IT world than other, mostly commer- tyofJava forcedbrowserdevelopersandthe again available free of charge for university cial, packages – a decision vindicated by Moodle community to block Java Applets instructors who wish to author online prob- subsequent developments. Launched and and cross-origin data, and we eventually lems for their students.[6] However, it can published in 2002 by Martin Dougiamas,[1] had to abandon JME. Fortunately, P. Ertl only be used within a cloud computing en- Moodle was designed as a free open source and P. Bienfait released a JavaScript suc- vironment, the servers being located in the e-learning environment and has since been cessor (JSME) under the open-source-type US. Only a few of the tools/algorithms used maintained and further developed by over BSD (Berkeley Software Distribution) li- by ACEorganic are in the public domain. 600 people from all over the world (our cense in time to prevent a forced shutdown Since we had neither the resources, course started with version 1.6, the current of our structure-based online exercises.[8] nor the skills, nor the time to program a version is 3.3) and now counts more than suitable e-learning environment such as 80’000 installations worldwide.[2] ACEorganic from scratch, we decided to 4. The Decision to Develop our own adopt a modular approach and use availa- 3.2 Structural Formula Editors and Tool to Draw Structural Formulae ble tools (Fig. 2, right) that would provide SMILES Generators and Convert them to Alphanumeric the following essential functionalities: We needed a simple but reliable draw- Strings 1. An e-learning environment with a built- ing tool that could be invoked via a link in in quiz module accepting and checking the Moodle course and would appear in a Under the impression of these events, alphanumeric strings as answers. separate window. Two of the editors avail- we nevertheless decided to develop our 2. An editor allowing structural formulae able in 2006/7 were considered: JME (Java own structural formula editor, mainly i) to be drawn on a computer or tablet. Molecular Editor) by P. Ertl, Novartis[7] to eliminate features that we consider un- 3. A tool generating an alphanumeric and MarvinSketch by ChemAxon.[5] As desirable in an educational context and ii) code that unequivocally reflects the Java Applets, both required the Java run- because of our plans to conduct future 1st drawn molecular structure and can be time environment on the student comput- year organic chemistry exams online and copied/pasted into the answer field of ers. Neither was open source but both had our uneasiness about using software with- the quiz. a documented API (application program- out knowing its source code. ChemiCal eduCation CHIMIA 2018, 72, No. 1/2 51

4.1 Generation of an Alphanumeric Technology) and the IUPAC (International when they are supposed to recognize it on Code from a Structural Formula Union of Pure and Applied Chemistry) their own? In 2006/7, the decision to use the to develop the InChI,[9] which uses the SMILES format to transfer structural in- McKay algorithm[13] for canonicalization formation from the formula editor to the and is biunique. The InChI generation code 5. MOSFECCS – A Newly Moodle quiz answer fields was a practi- is open source, but the testing suite for pro- Developed Structural Formula cal one: JME produced SMILES codes grammers who want to incorporate it in- Editor and SMILES Generator anyway, and the InChI (International to their own software is commercial. The Chemical Identifier),[9] now the IUPAC- NCI-NIH (National Cancer Institute of the The drawing tool and SMILES gener- recommended format, was not yet broadly National Institutes of Health) website[14] ator MOSFECCS (MOlecular Structural accepted and consists of rather long and in- offers a tool translating any formally cor- Formula Editor and Calculator of comprehensible strings, even for relatively rect SMILES code into the corresponding Canonical Smiles, Fig. 4), which we in- small molecules. SMILES codes, on the Daylight-SMILES. troduced in 2016, follows a more didactic other hand, can be straightforwardly inter- perspective, tolerating many kinds of mis- preted, i.e. re-transformed into a molecular takes and treating each resonance structure structure by the human mind (Fig. 3). 4.3 Editor Design and Features: or tautomer as a distinctive species. Conflict between Didactic Goals and MOSFECCS generates biunique 4.2 SMILES (Simplified Molecular Efficiency for Database Searches SMILES codes that are, as far as tested, Input Line Entry Specification)[10] JME/JSME, MarvinSketch and sim- identical to Daylight-SMILES produced In 1988, D. Weininger published the ilar editors are now standard input tools by the NCI-NIH translator, except in cas- rules and algorithms of SMILES, a pro- for structure-based database searches. For es where we deliberately deviate from the cedure for encoding the information of reasons of convenience, they auto-correct Daylight-SMILES convention. This ap- a structural formula as an alphanumer- common errors such as pentavalent carbon, plies to rings with a maximum number of ic string.[10b] This encoding, however, is automatically generate requisite formal conjugated double bonds: Instead of trying not biunique: A given structural formu- charges and offer palettes with ready-made to assign aromaticity (lower-case letters are la can usually be encoded into different functional groups etc. With the objective of used by Daylight-SMILES for atoms in ar- formally correct SMILES which, upon finding a maximum if not all of the entries omatic rings), MOSFECCS uses small let- back-translation, will all give the same for a given substance in a database, mod- ters for all atoms of even-membered man- structural formula. The popular program ern structure editors try to merge different cude rings (see Fig. 4).[15] This has nothing ChemDraw,[11] for example, generates for- representations such as separate resonance to do with the concept of aromaticity but mally correct SMILES but not unique ones structures and even tautomers. Daylight- simply avoids having to program all their and converts formally correct SMILES to SMILES tries to recognize and assign aro- trivial resonance or double-bond-shifted a correct structural formula (see Fig. 3). maticity for this purpose, whereas InChI on- structures as correct answers in a quiz (each This missing one-to-one correspondence ly uses connections and does not distinguish unsymmetrically substituted benzene ring between standard SMILES and structur- between bonds of different bond order. in a molecule would multiply the number al formulae makes the former unsuitable All these features are inappropriate for of required entries by two). For the pur- for database searches. Similarly, using a formula editor used to teach first-year poses relevant in a basic organic chemistry non-biunique SMILES as input in Moodle organic chemistry, where the formalism course, there are no negative consequences quizzes would require all possible for- and rules of classical structure theory or if the SMILES code ignores the position mally correct SMILES codes for a given electron delocalization are key subject of alternating double bonds in even-mem- structure to be stored as correct answers, matters: the students must be able to make bered mancude rings. For other unsaturat- which is impossible. To be usable in such mistakes and learn from them! Online ex- ed even-membered rings (e.g. σ-complex- a context, the alphanumeric encoding has ercises, let alone exams, are practically es) and for odd-membered unsaturated to be biunique: a given structural formula worthless if the used structure editor auto- rings (e.g. pyrrole, cyclopentadienyl anion, must be transformed into a single, unique corrects errors and gives out warnings if tropylium cation), the generated SMILES SMILES, regardless of the sequence in valences are exceeded or formal charg- code reflects only the particular resonance which atoms and bonds were drawn. This es forgotten. What is the point of a tool structure drawn. difficulty, known in graph theory as the that transforms all resonance structures Stereochemistry is a pivotal subject in ‘canonicalization’ or ‘unique numbering’ or tautomers into a single one or tells the the first year of an organic chemistry course. problem, was addressed by Weininger et students that a particular cycle is aromatic Following the so-called isomeric-SMILES al. in a second paper,[10c] which describes a procedure to generate ‘unique SMILES’. However, others found examples where the procedure of ref. [10c] failed and led to different ‘unique’ SMILES for different standard SMILES isomeric SMILES absolute SMILES drawing sequences.[12] In the meantime, not biunique +stereochemical +stereochemical the rights to SMILESTM had been trans- speciﬁcations speciﬁcations CCC(C)C=CC +canonicalization ferred to Daylight Inc. and, although this C/C=C/[C@@H](C)CC company claimed to have eliminated the CC(CC)C=CC CC[C@H](C)\C=C\C (2E,4S)-4-methylhex-2-ene CC=CC(CC)C generated by generated by problems, and that the so-called Daylight- JSME, ChemDraw MOSFECCS, NCI translator SMILES were now truly biunique, neither (Daylight®-SMILES) the corresponding changes in the algo- InChI code: 1S/C7H14/c1-4-6-7(3)5-2/h4,6-7H,5H2,1-3H3/b6-4+/t7-/m0/s1 rithm nor the actually used code can be peer-reviewed, because they are not in the Fig. 3. Encoding of structural formulae into alphanumeric strings. Different SMILES conventions, public domain. This unsatisfactory situa- the InChI code, and output of SMILES generators used by different structural formula editors with tion was one of the reasons prompting the (2E,4S)-4-methylhex-2-ene as example. To be useful for online quizzes and tests, the encoding NIST (National Institute of Standards and must be biunique (red). 52 CHIMIA 2018, 72, No. 1/2 ChemiCal eduCation

Fig. 4. MOSFECCS of incorrect or partially correct answers (MOlecular Structural (Fig. 5). Human teachers can often recog- Formula Editor nize the error in reasoning which leads to and Calculator of a particular wrong answer and distinguish Canonical SMILES): it from simple guessing or a slip of the View of the canvas with the drawing pen. When wrong answers that are relat- toolboxes at the up- ed to foreseeable errors in reasoning are per and left edges. submitted in online quizzes, the students The currently active can be given specific feedbacks with tips tools are marked enabling them to take the next hurdle and with a red bar. The improve in the next trial. Actually, the inset shows the teacher can identify the most frequent mis- SMILES code ‘[NH3+] takes by looking at the statistics of given CCCC[C@H]([NH3+]) answers and improve the feedback accord- C([O-])=O.CC(=O) ingly over time. One problem, however, Oc1ccccc1C([O-])=O’ generated for the persists: computers cannot distinguish drawn formula (salt between faulty reasoning and slips of the from acetylsalicylic pen. Whenever the latter add on top of the acid and l-lysine) former (and this happens more often than that can be copied one would believe), a useful feedback is and pasted into the not possible. answer field of a quiz (not shown). 6.2 Online Exams: The Grading of Partially Correct Answers For authors of online exams, automatic grading of partially correct answers that include anticipatable mistakes is a particular challenge. In a nomenclature question, for (integral component of Daylight-SMILES) tures on the canvas. For online exercises example, one would normally assign a par- convention, MOSFECCS uses the symbols this is inexpedient, because it requires all tial grade if the systematic name submitted @ and @@ to specify the absolute configu- permutations of individual, dot-separated for a given enantiomer is correct, but the ste- ration of stereogenic centres and the marks sub-strings to be programmed as correct reodescriptor is either wrong or missing. If // and /\ for the configuration of double answers in the quiz module. To eliminate the molecule contains a single stereogenic bond units (Fig. 4).[16] In the case of stereo- this nuisance, MOSFECCS always concat- centre only, this situation is easily handled genic centres with an implicit hydrogen enates in a well-defined order by arranging by an additional entry in the list of expect- or a lone pair as one of the four ligands, the individual SMILES sub-strings accord- ed answers. If, however, it includes more the 3D-coordinates of this ‘invisible’ li- ing to decreasing length and – if the length than one stereogenic centre, the number of gand have to be deduced from the way the is the same – in alphabetic order (Fig. 4). extra entries in the answer list grows very bonds to the three visible ones are drawn. The scripting part of MOSFECCS is rapidly. One way to reduce the list of an- As analysed comprehensively in the tech- written entirely in JavaScript (ECMA script swers considered for partial grading is the nical InChI manual,[17] only a subset of all 5.1) and uses Canvas elements as defined use of wildcards (similar to UNIX shells, possible ways to draw a centre with tetrahe- in HTML 5 for interactive drawing. So far, Moodle interprets the asterisk * as ≥0 arbi- dral or trigonal-pyramidal bond geometry it has been tested for the current versions trary characters). The text-based question defines its absolute configuration unambig- of the browsers Firefox (Windows 10, types of Moodle, i.e. SHORTANSWER uously. If MOSFECCS encounters equivo- MacOS, iOS, Android), Safari (MacOS, and CLOZE, accept wildcards in the pro- cally drawn bond geometries at stereogenic iOS), Edge (Windows 10), Chrome grammed answers and correctly match centres during the SMILES generation, it (MacOS, Android) and the Safe Exam these against the student’s submitted an- marks them on the canvas with a red square Browser[18] (MacOS and Windows 10). In swer. Regular expression (regex) matching, and issues a warning instead of a SMILES recent years, more and more students use known from the grep command in UNIX code, thus forcing the student to improve tablets rather than laptop or desktop com- shells or from programming languages the drawing until it is unambiguous. puters. Sketching a structural formula with such as Perl or PHP, would be much more When the answer to a quiz encom- fingertips or a stylus instead of a mouse powerful than a simple wildcard. Regex passes several structures (question type: requires some adjustment for any drawing modules for Moodle have indeed been de- “draw the three best resonance structures program. Currently, MOSFECCS runs on veloped, but unfortunately they are not part of prop-2-enal”), it is convenient to draw iOS-, Android-, and Windows 10-based of the Moodle core distribution and have them all at once on the canvas and submit tablets. to be installed from the third party plugin them together to the SMILES generator repository.[19] This means that their oper- (Fig. 4). The SMILES convention allows ability may be lost after the next Moodle for specification of multiple structures by 6. A Particular Challenge: upgrade, which is, alas, a strong argument a single string in which the codes for the The Handling of Partially Correct against using regex in the course. individual molecules are simply concate- Answers in an Online Test The incorporation of wildcards in nated with a period as separator. However, SMILES strings is of limited usefulness. the convention includes no rule about the 6.1 Feedback to Wrong or Partially Due to the process of canonicalization, order in which the individual sub-strings Correct Answers small changes in a structure, such as the ex- should be linked together; with most ed- Online problems are most valuable if change of a methyl for an ethyl substituent itors, the sequence depends on the order they provide a feedback which is specific or a different substitution pattern, may lead of drawing or the position of the struc- to the mistake the student made in the case to a completely different sequence of atom ChemiCal eduCation CHIMIA 2018, 72, No. 1/2 53

alphanumeric code as implemented (via MarvinSketch) in the ACEorganic e-learning environment.

8. Conclusions

After setting up and running a 1st year organic chemistry Moodle course for a dec- ade, we are pleased to see that this supplementary class module is becoming increasingly popular among the students. This is partly due to the fact that, with the advent of tablet computers and smartphones, students are getting used to the pervasive use of electronic tools and resources. But it is also related to the refinement of the course material, in particular the multiplication of the online quizzes which now cover a large range of topics. We learnt the lesson that while an elaborate course with online problems allows a very convenient training of students at almost no cost per capita, the effort and time required for the setup and maintenance are substantial. Since the didactic value of online quizzes crucially depends on the quality of the feedbacks, experienced teachers are needed for the design and elaboration, whereas the tech- Fig. 5. Sample quiz question from our Moodle Course Organic Chemistry I: Transcription of a nical editing and testing can be delegated given Fischer into a Natta projection. The structural formula editor MOSFECCS is used by the stu- to others. dent to draw the Natta projection and generate its SMILES code which is copied/pasted into the The initial choice of e-learning envi- answer field (‘Antwort’) (left). In the shown example, the proposed structure and, by implication, ronment, core software tools, structural the input SMILES code contains two mistakes. After this partially wrong answer is submitted by formula editor, and the interface to con- pressing the check button (‘Prüfen’), Moodle returns a specific feedback saying that two stereo- vert the formulae to alphanumeric data has genic centres have the wrong configuration. The student can now revise the answer and resubmit. far-reaching consequences and should be After working through all the questions of a test, the student receives a general feedback includ- well considered. For a course with more ing the correct answers as well as any explanations elaborated by the teacher (right). than a thousand quiz questions and an even greater number of answers/feedbacks, the decision to switch to other software com- symbols inside the SMILES string. On the 7. Current Limitations and Possible ponents or standards requires a substantial other hand, when the correct answer to a Extensions of MOSFECCS part of the work already invested to be re- quiz question is a multi-structure SMILES, done. For the same reason, it is strongly wildcards do allow to check for the pres- The current version of our structur- advised to resist the temptation of using ence of one or several of the constituent al formula editor and SMILES generator new, fancy question types that offer tools sub-codes within the aggregate SMILES (MOSFECCS version 2) is tailored for such as an integrated molecular formula string. This allows for automatic grading organic chemistry. To make it useful for editor[22] or the use of powerful regular ex- when a student finds only two out of a total inorganic chemistry as well, other types pressions. Such tools periodically appear of three isomers or resonance structures. of bonds such as coordinative bonds, in the Moodle world, but they usually have In view of the above shortcomings, it π-coordination, etc. and stereoisomerism a limited lifetime because they are not in- is indispensable that, after an online ex- at centres with square-planar, trigonal-bi- corporated in the Moodle core modules. am, the teacher carefully reviews the sub- pyramidal and octahedral bond geome- The history of our project also reveals mitted answers, notably those marked as try would have to be included (the latter the problems associated with a dependence wrong or partially correct by the computer. type is already defined in the Daylight- on proprietary software or on software as Fortunately, Moodle makes this perusal of SMILES convention).[16] For questions a service (SaaS, cf. ACEorganic).[6] In the students’ answers easy and allows for involving chemical reaction schemes, it the world of scientific computing, there manual modification of the grades. For an would be desirable to extend the alphanu- are many examples of academic authors efficient and reliable human scrutiny of meric encoding in order to include reaction having created an ingenious software tool submissions consisting of SMILES codes, arrows and specifications for reactant, re- which, at some point, was commercialized however, these have to be re-transformed agent/catalyst and product, similar to the without publication of the code. The ensu- into structural formulae. For this purpose, reaction-SMILES notation published by ing trouble is not primarily of pecuniary we have developed a script that imports the Daylight.[21] Questions expecting the stu- nature and many copyright owners even SMILES answers of all candidates into a dents to formulate reaction mechanisms allow a free non-commercial use of their spreadsheet. With the ChemDraw plugin would require the optional drawing of software (Marvin, JME, JSME). The main for Excel[20] the codes can then be automat- lone pairs and curved arrows in the struc- inconvenience is the lack of control over ically re-converted to structural formulae tural formula editor as well as the encod- one’s most important tools and the nonex- for visual inspection. ing of the electron-shift notation in the istent guarantee that proprietary software 54 CHIMIA 2018, 72, No. 1/2 ChemiCal eduCation

will be maintained and available over time. [1] M. Dougiamas, P. C. Taylor, ‘Interpretive ana- [14] https://cactus.nci.nih.gov/translate, accessed In the context of online chemistry exams, lysis of an internet-based course constructed 29 Sept. 2017. undisclosed source code may lead to an ad- using a new courseware tool called Moodle’, [15] IUPAC, ‘Compendium of Chemical Termino- Proceedings of the Higher Education Research logy’, 2nd ed. (the ‘Gold Book’), compiled by ditional complication: What is the proper and Development Society of Australasia A. D. McNaught and A. Wilkinson, Blackwell response to the possible appeal of a student (HERDSA) 2002 Conference, Perth, Western Scientific Publications, Oxford, 1997. XML claiming to have drawn the correct molec- Australia. on-line corrected version: http://goldbook. ular structure but that a bug in the formula [2] https://moodle.org, accessed 29 Sept. 2017. iupac.org (2006-) created by M. Nic, J. Jirat, B. Kosata; updates compiled by A. Jenkins. editor caused an error in the encoding? [3] For a project to directly interpret bitmap graph- ics of scanned structural formulae see: I. V. ISBN 0-9678550-9-8. https://doi.org/10.1351/ Considering these risks, we feel that Filippov, M. C. Bicklaus, J. Chem. Inf. Model. goldbook, accessed 29 Sept. 2017. Last update: the only satisfactory solution is the use 2009, 49, 740. 2014-02-24; version: 2.3.3. DOI of this term: of open-source software throughout. [4] R. R. Chamala, R. Ciochina, R. B. Grossman, https://doi.org/10.1351/goldbook.M03695. Accordingly, we plan to make the structur- R. A. Finkel, S. Kannan, P. Ramachandran, J. [16] Daylight Theory Manual, Version 4.9, 2011, Chem. Educ. 2006, 83, 164. Daylight Chemical Information Systems, Inc. al formula editor MOSFECCS available [5] https://www.chemaxon.com, accessed 29 Sept. Laguna Niguel, CA, http://daylight.com/day- under an open-source license as soon as its 2017. html/doc/theory/theory.smiles.html, accessed development and initial round of testing [6] https://ace.chem.illinois.edu/ace/public/fea- 29 Sept. 2017. have been completed. tures/index.html, accessed 29 Sept. 2017. [17] S. E. Stein, S. R. Heller, D. V. Tchekhovskoi, I. Pletnev, IUPAC International Chemical Overall, our setting up of Moodle [7] P. Ertl, J. Cheminf. 2010, 2, 1; see also: http:// www.molinspiration.com/jme/index.html, ac- Identifier (InChI) InChI version 1, software ver- courses with online exercises, although cessed 29 Sept. 2017. sion 1.04, 2011, Technical Manual. quite operose, was and continues to be a [8] B. Bienfait, P. Ertl, J. Cheminf. 2013, 5, 24. [18] http://safeexambrowser.org, accessed 29 Sept. rewarding and instructive experience. As [9] S. R. Heller, A. McNaught, I. Pletnev, S. E. Stein, 2017. a result, the students have at their dispos- D. V. Tchekhovskoi, J. Cheminf. 2015, 7, 23. [19] https://docs.moodle.org/33/en/Regular_ [10] Concepts for encoding the information con- Expression_Short-Answer_question_type, ac- al a considerable and ever-growing pool tained in a structural chemical formula as an cessed 29 Sept. 2017. of online quizzes, which they can use ad alphanumeric string – more amenable to elec- [20] The ChemDraw for Excel plugin (included in libitum. We – and possibly the students tronic data processing than a picture – have a Chem Office 16, Perkin Elmer Inc.) is only too – are now anxious to take the next step, long history, ranging from a) the Wiswesser available for Microsoft Windows and requires the 32bit version of Excel (Microsoft Office). namely the implementation of online ex- Line Notation: W. J. Wiswesser, J. Chem. Inf. Comput. Sci. 1982, 22, 88; over b) SMILES [21] A. R. Leach, J. Bradshaw, D. V. S. Green, M. ams including the submission of structural (Simplified Molecular Input Line Entry M. Hann, J. Chem. Inf. Comput. Sci. 1999, 39, formulae as answers. Specification): D. Weininger, J. Chem. Inf. 1161. Comput. Sci. 1988, 28, 31; and c) its variants, [22] https://docs.moodle.org/33/en/question/type/ Acknowledgements the isomeric-SMILES, unique-SMILES: D. jme, accessed 29 Sept. 2017. We gratefully acknowledge Markus Beck Weininger, A. Weininger, J. L. Weininger, for contributing many new online exercises in J. Chem. Inf. Comput. Sci. 1989, 29, 97; d) Daylight-SMILESTM: http://www.daylight.com/ 2008, the LET team (ETH Zurich) for support- smiles/index.html, accessed 29 Sept. 2017; ing us from the beginning and maintaining the e) and open-SMILES: http://opensmiles.org, Moodle installation at the school, ETH Zurich accessed 29 Sept. 2017; Specification: http:// for financial support (FILEP grant, 2008) and opensmiles.org/opensmiles.html, accessed the many teaching assistants as well as a number 29 Sept. 2017); to the InChI (International of students of the courses Organic Chemistry I Chemical Identifier) code: see ref. [9]. and II for students of Biology, Pharmaceutical [11] ChemDraw® and ChemOffice® 16, Perkin Sciences, and Health Sciences and Technology Elmer Inc. (ETH Zurich), who contributed ideas or report- [12] G. Neglur, R. L. Grossman, B. Liu, in ‘Data Integration in the Life Sciences’, Proceedings ed bugs, thereby helping to improve the online of the Second International Workshop DILS, course. We thank Dr. Peter Ertl (Novartis) for 2005, Eds. B. Ludäscher, L. Raschid, pp 145- letting us use JME and JSME. 157, Springer, 2005. [13] a) B. D. McKay, Congr. Numer. 1980, 30, 45; b) B. D. McKay, A. Piperno, J. Symb. Comput. Received: October 7, 2017 2014, 60, 94.