Supporting Information Reker et al. 10.1073/pnas.1320001111 SI Materials and Methods Library of Pharmacologically Active Compounds Off-Target Prediction. All starting materials and solvents were purchased from Sigma- The library of pharmacologically active compound (LOPAC) col- Aldrich, Fluka, or ABCR and were used without further purification. lection (version 1280) was downloaded (www.sigmaaldrich.com) Microwave syntheses were performed on a Biotage Initiator and preprocessed as described in the main text. Previously pre- microwave synthesizer. computed self-organizing maps (SOMs) clustering the whole col- Melting points (mp) were recorded on a Büchi M560 apparatus lection of bioactive reference analogs (COBRA, inSili.com LLC, and are uncorrected. Proton and carbon NMR (1H and 13C Zurich) were used to project queries onto clusters of reference NMR) spectra were recorded on a Bruker Avance 400 (400 and compounds. Tanimoto similarities (Tc) of every query were 100 MHz, respectively). All chemical shifts are quoted on the δ computed from Dayllight fingerprints of coclustered drugs. When scale in parts per million using residual solvent peaks as the a prediction for a particular target used at least one reference with internal standard. Coupling constants (J) are reported in hertz aTc> 0.2, the prediction was discarded. This procedure left us with the following splitting abbreviations: s, singlet; d, doublet; with a list of predictions using structurally unrelated references. dd, doublet of doublets; td, triplet of doublets; m, multiplet. Filtering for high-confidence predictions (P < 0.001) left us with Analytical HPLC-MS was carried out in a Shimadzu LC-MS2020 the final set of predictions. system, equipped with a Nucleodur C18 HTec column, under an + appropriate gradient of acetonitrile: H2O( 0.1% formic acid in Screening Assays. Activity tests were performed by contractors on each solvent), and a total flow rate of 0.5 mL/min. Preparative a fee-for-service basis. EC50 values against the B1 receptor were HPLC was carried out on a Shimadzu LC-8A system, coupled to determined by measuring intracellular calcium by fluorimetry in – a Nucleodur 100 5C18 HTec column, and a SPD-20A UV/Vis human recombinant CHO cells at Cerep (Ref. 1109) (1). The detector. High-resolution MS (HRMS) analyses were performed agonist effect on vanilloid 1 (TRPV1) was determined by mea- on a Bruker Daltonics maXis electrospray ionization (ESI) quad- suring intracellular calcium by fluorimetry in human recombi- rupole time-of-flight device. MS analyses were operated in posi- nant CHO cells at Cerep (Ref. 1640) (2). The antagonist effect m/z tive-ion mode with ESI. Nominal and exact values are reported on neurokinin 1 (NK1) was determined by measuring intracellular ≥ in daltons. All compounds present purity 95% based on LC-MS calcium by fluorimetry in U373 cells at Cerep (Ref. 2192) (3). analysis. Screening against proteases was carried out at Reaction Biology Corp. using suitable assay kits. Docking. Docking was performed using GOLD (version 5.2) with HIV1 protease cocrystalized with the de novo design template 1 Spectral Data for Compound 1. H NMR (CD3OD, 400.13 MHz) amprenavir [Protein Data Bank (PDB) code 1HPV]. The H – H 1 2 d 1.36 (9H, s, (C 3)3), 1.91 2.02 (2H, m, C 2), 2.43 (3H, s, binding site was defined by the position of amprenavir. and CH ), 2.62 (1H, dd, J = 13.6 and 9.7 Hz, CH ), 2.80 (1H, dd, J = were preprocessed using the Molecular Operating Environment 3 2 13.6 and 5.2 Hz, CH ), 3.35–3.52 (4H, m, CH ), 3.74 (1H, d, J = (MOE; Chemical Computing Group, Montreal) wash function, 2 2 5.0 Hz, CH), 3.94 (1H, m, CH), 5.03 (1H, m, CH), 6.50 (1H, d, and starting structures were calculated using the MOE energy J = 9.4 Hz, NH), 7.16–7.27 (5H, m, Ar-H), 7.42 (2H, d, J = 8.0 minimizer. Hydrogen atoms were added to the protein, and waters H J = H 13 were deleted using internal GOLD functions. Atom typing for Hz, Ar- ), 7.73 (2H, d, 8.0 Hz, Ar- ); C NMR (CD3OD, ligand and protein was activated, and conformations were scored 100.61 MHz) d 21.55; 28.74; 32.28; 37.00; 47.49; 54.55; 56.12; using the ChemPLP scoring function. 73.86; 75.65; 80.23; 127.37; 128.79; 129.28; 130.55; 130.64; 131.02; 134.78; 139.48; 145.38; 173.53; HRMS-ESI calc. (C26H34N2O7S + + Repurposing. The publicly available target prediction methods Na ): 541.1979, found: 541.1976. Similarity Ensemble Approach (SEA; http://sea.bkslab.org/; 1 version ChEMBL 10), Prediction of Activity Spectra (PASS; Spectral Data for Compound 2. H NMR (CD3OD, 400.13 MHz) – H H J = www.pharmaexpert.ru/passonline/), Semantic Link Association d 1.63 1.80 (2H, m, C 2), 2.32 (3H, s, C 3), 2.76 (1H, dd, H J = H Prediction (SLAP; http://cheminfov.informatics.indiana.edu:8080/slap/), 14.0 and 7.0 Hz, C 2), 2.85 (1H, dd, 14.0 and 7.0 Hz, C 2), J = H H and the method SuperPRED (http://bioinformatics.charite.de/ 3.16 (2H, dd, 9.0 and 5.2 Hz, C 2), 3.24 (2H, m, C 2), 3.73 superpred/) were applied using their freely available online serv- (1H, td, J = 7.0 and 3.0 Hz, CH), 4.17 (1H, d, J = 3.0 Hz, CH), ices. Compounds 1 and 2 and the design template amprenavir were 4.74 (1H, m, CH), 7.11–7.26 (5H, m, Ar-H), 7.31 (2H, d, J = 8.2 H J = H 13 provided as simplified molecular input line entry system (SMILES). Hz, Ar- ), 7.60 (2H, d, 8.2 Hz, Ar- ); C NMR (CD3OD, Results were manually extracted and sorted according to provided 100.61 MHz) d 21.50, 32.06, 34.51, 47.20, 54.42, 56.07, 70.56, confidence scores (scores or P values). For the PASS method, which 76.37, 128.66, 128.77, 130.02, 130.73, 130.98, 134.71, 136.50, 145.50, + provides two orthogonal confidence scores, P(active) and P(in- 171.37; HRMS-ESI calc. (C21H26N2O5S + H ): 419.1635, found: active), the ratio of these scores P(active)/P(inactive) was used. 419.1634. 1. Simpson PB, Woollacott AJ, Hill RG, Seabrook GR (2000) Functional characterization of 3. Eistetter HR, et al. (1992) Functional characterization of neurokinin-1 receptors on bradykinin analogues on recombinant human bradykinin B(1) and B(2) receptors. Eur J human U373MG astrocytoma cells. Glia 6(2):89–95. Pharmacol 392(1-2):1–9. 2. Phelps PT, Anthes JC, Correll CC (2005) Cloning and functional characterization of dog transient receptor potential vanilloid receptor-1 (TRPV1). Eur J Pharmacol 513(1-2):57–66. Reker et al. www.pnas.org/cgi/content/short/1320001111 1of5 A B C D Ile Ile H2O Asp25 Asp25 Asp29 Asp Asp Gly Fig. S1. Superimposition of the top three docking poses for 1 (A) and 2 (B). Docking poses of compounds 1 (C) and 2 (D) in the catalytic site of HIV-1 protease (PDB code 1HPV) (1). Hydrogen bonds with Asp25 residues are highlighted. Images were generated with PyMOL v1.4.1 (Schrödinger), and for clarity, only polar hydrogen atoms are shown. 1. Kim EE, et al. (1995) Crystal structure of HIV-1 protease in complex with VX-478, a potent and orally bioavailable inhibitor of the enzyme. J Am Chem Soc 117(3):1181–1182. Fig. S2. (A) Performance of the prediction models using solely the chemically advanced template search (CATS) or the MOE representation, measured in percentage of molecules for which the correct target is predicted with significant confidence (P < 5%). Each point represents the performance on molecules associated with one particular target. (B) ROC curves for the different combination functions (arithmetic average, red; maximum, gray; minimum, blue; geometric average, dashed green; CATS2 only, pink; MOE only, dashed black). (C) Structures of fenofibrate and its nearest neighbor from the training data annotated to block sodium ion channels and the concentration-response curve of fenofibrate against Nav1.5 ion channels (n = 2, mean and SEM). Reker et al. www.pnas.org/cgi/content/short/1320001111 2of5 Reker et al. www.pnas.org/cgi/content/short/1320001111 B A Fig. S3. 7.608 7.737 180 8.0 7.716 7.604 8.0 7.426 j &k 173.53 7.406 7.599 170 2.00 7.269 7.5 170 7.587 171.37 7.250 1 2.03 7.233 160 H NMR, 7.583 l, m&n 2.15 7.196 7.5 2.92 2.00 7.578 7.0 7.181 150 7.162 160 7.314 6.513 2.05 7.312 145.38 6.490 6.5 140 5.030 5.01 0.60 7.293 139.48 3.961 7.0 13 134.78 3.949 7.254 131.02 150 130 C NMR, and 130.64 3.944 7.250 130.55 6.0 129.28 3.937 128.79 3.932 7.238 127.37 120 3.925 145.50 6.5 7.233 3.914 5.5 140 3.751 7.229 110 3.738 7.218 3.516 d 136.50 5.0 3.504 100 0.81 134.71 7.214 3.485 6.0 130.98 3.474 130 7.193 1 130.73 3.459 H- 90 4.5 130.02 7.189 3.452 128.77 3.435 1 7.186 H COSY data for compound 3.429 128.66 80 5.5 7.172 80.23 3.413 4.0 120 b 3.407 7.151 75.65 73.86 0.98 3.394 70 7.147 a 3.363 0.83 3.353 3.5 7.142 g &h 2.823 60 5.0 3.16 110 2.811 7.131 56.12 0.76 2.788 54.55 4.168 3.0 2.776 50 2.644 4.161 47.49 2.620 5.75 0.87 c 2.609 100 40 4.5 3.752 0.88 2.5 2.585 3.745 37.00 2.428 3.00 f 2.029 3.734 30 32.28 28.74 2.019 3.727 2.0 2.007 1.04 i 0.93 90 4.0 1.997 20 21.55 0.99 3.715 1.994 1.985 3.708 1.5 10 1.982 3.238 1.973 1.01 7.96 e 1.960 80 3.228 3.5 ppm 1.919 1.0 1 3.217 1.913 1.906 ( 76.37 A 3.213 pp 1.889 ) and 1.93 3.209 m 1.359 70 70.56 1.93 3.0 3.205 l, m&n 3.201 1 2.01 3.177 H NMR and 60 3.164 G16-20130424_TR62C 31X:\data\pgsnmr\nmr 2.5 3.155 56.07 3.141 54.42 3.02 7 50 2.872 2.0 2.853 47.20 2.837 c 2.818 HN 13 40 O b 1.01 6 C NMR data for compound 2.784 1.5 NH a 0.95 OH 2.767 O OH O 34.51 2 2.748 30 32.06 O 2.731 O 5 e O 1.0 d 2.322 i 2.055 g h 20 1.804 21.50 N O 1.791 4 N S 0.5 ppm O 1.768 O S 1.755 O ppm 1.635 f 1.631 3 j &k 2 2 ( F2 [ppm] B ).