Chemically Advanced Template Search (CATS) for Scaffold‐Hopping and Prospective Target Prediction for ‘Orphan’ Molecules

Abstract

Drug discovery is driven by the identification of new chemical entities (NCEs).1,2 Virtual screening and de novo design techniques have been proven to serve this purpose, thereby complementing experimental biochemical and biological approaches.3 Still, it remains a matter of debate, which particular molecular representation and similarity index are preferable for a given drug target in order to identify appropriate NCEs with minimal synthetic and testing effort involved.4 Ligand-based chemical similarity approaches have been effectively applied to large-scale activity and target prediction for known drugs, some of the prominent methods being PASS developed by Poroikov et al.,5 the techniques conceived by Mestres and co-workers,6 and the similarity ensemble approach (SEA) implemented by the Shoichet group.7 Here, we compared several popular two-dimensional molecular representations for their ability to retrieve actives (enrichment potential) and chemotypes (scaffold-hopping potential) from a collection of druglike bioactive compounds. Subsequently the applied chemical advanced template search (CATS)8 was applied to predicting potential drug targets for a virtually assembled combinatorial compound library, from which we synthesized and successfully tested candidate compounds. The results demonstrate that CATS is not only suited for its intended purpose of NCE retrieval by scaffold-hopping,9 but also for reliable target profiling of 'orphan' virtual molecules.10 It thereby complements the suite of available validated tools for target prediction. A framework for retrospective evaluation of similarity searching runs with different molecular representations ('descriptors') was established on basis of the COBRA collection of druglike bioactive compounds,11 employing Euclidean distances for metric descriptors and the Tanimoto coefficient for fingerprint descriptors.12 COBRA contains 12 642 manually curated entries with 980 target protein subtype annotations. For 170 macromolecular drug targets with a minimum of 20 annotated active ligands per target, each compound annotated as 'active' was selected as a query in turn, and compared to all remaining compounds in the screening pool in terms of molecular descriptor similarity, finally yielding sorted results lists with the most similar or least distant pool compounds sorted to the top. Although there are large collections of bioactive compounds available in the public domain,13 we used the carefully compiled COBRA collection to i) reduce the risk of erroneous activity data and faulty compound structures,14 and ii) avoid redundancy with existing tools that are based on such public structure-activity data. In addition, we intend to probe the value of a comparably small but well curated reference compound pool for target prediction. We used a representative set of descriptors and fingerprints for benchmarking. 'Morgan' fingerprints, closely related to extended-connectivity fingerprints (ECFP), are based on radial assessment of non-predefined potentially infinite molecular fragments.17 The 'AtomPair' descriptor can be seen as a CATS predecessor merely denoting the occurrence of all pairs of atoms at a given topological distance.18 The 'MACCS' keys represent substructure-based fingerprints,19 and the 'RDkit' fingerprint implements a Daylight-like fingerprint based on hashed molecular subgraphs.20 Latter fingerprints and descriptors were calculated using the open-source software package RDkit.21 Finally, the 'MOE2D' descriptor consists of a standardized vector of physicochemical properties provided by the Molecular Operating Environment (v2011, Chemical Computing Group, Montreal). At this point, we analyzed two versions of CATS vectors, namely the originally described CATS18 and CATS2, which distinguishes lipophilic from aromatic atoms during typing, thereby resulting in more pharmacophore type pairs and consequently a higher dimensionality of the descriptor than CATS1, which lacks the aromatic atom type. For both descriptors we employed 'types scaling', which mitigates the potential dominance of prevalent pharmacophore feature types, and a maximal correlation distance of 10 bonds.22 An example of CATS descriptor calculation is presented in Figure 1. Principle of CATS descriptor calculation. The molecular structure (Step 1) is reduced to the molecular graph, and feature types are assigned (Step 2; L, lipophilic; R, aromatic; A, hydrogen-bond acceptor; D, hydrogen-bond donor). Then, atom pairs for all feature pairs are counted (Step 3), and the final descriptor values are scaled (Step 4). Here, the raw values were divided by the respective λ value (sum of atom type pair occurrences). Note that not all vertices in the molecular graph are considered 'pharmacophoric'. These possess no feature types. We employed the Receiver Operating Characteristic (ROC) related BEDROC score for actives-retrieval benchmarking.15 For our study, the alpha level of the BEDROC method was set to 160.9, which corresponds to the top 1 % of the screening list contributing 80 % of the score. Murcko scaffold16 diversity among the set of actives within the top 1 % of respective screening lists served as measure for scaffold-hopping potential. Albeit state-of-the-art radial fingerprints and atom-pair fingerprints outperformed CATS descriptors in terms of the number of actives retrieved (Figure 2A), the latter ratify their intent of design by delivering the overall highest ratio of diverse scaffolds among retrieved actives. Scaffold-hopping potential was determined by examining the distribution of relative scaffold diversities r, which is the ratio of differing scaffolds s to the number of retrieved actives n among the top 1 % of respective screening runs. While s correlates to the BEDROC scores when comparing different descriptors, r unveils the CATS1 descriptor as the most suitable descriptor for scaffold-hopping among the compared molecular representations (Figure 2B). In terms of BEDROC scores estimating the enrichment potential, radial fingerprints (Morgan) and Carhart-type atom pairs (AtomPair) performed similar, as did the CATS2 and MOE descriptors, while MACCS, CATS1, and RDkit fingerprints formed a third group (Figure 3A). With respect to scaffold-hopping potential, the groups vary, with CATS1 and MOE2D pairing up, as well as CATS2 and MACCS (Figure 3B). It might thus be advisable to select one method from each group for similarity searching and compare ranked results lists, e.g. by data fusion.23 We wish to point out that the grouping of methods depicted in Figure 3 should be treated with caution, as the dendrograms are likely to vary for other reference data sets and chemotype/target coverage. Comparison of molecular representations for their abilities to retrieve known actives (A) and scaffolds (B) from a collection of druglike bioactive compounds (COBRA). Violin plots show the shapes (gray), medians (white circle) and quartiles (thick lines) of the distributions. Similarity of molecular representations in terms of their enrichment (A) and scaffold-hopping potential (B). Pair-wise, one-sided Wilcoxon rank sum tests24 were performed for the BEDROC score distributions of the descriptors. Clustering the obtained p-values with Ward's method25 resulted in the depicted dendrograms. The outcome of this limited benchmark study is in agreement with a large-scale systematic analysis of 2D fingerprint methods by Sherman and co-workers, who conclude (…︁) if the objective of a screen is to identify novel, diverse hits, then a less specific atom-typing scheme may be more appropriate.26 The CATS representation of molecular graphs and pharmacophoric features serves this purpose of finding new chemotypes. When using the descriptor, one should not expect highest possible enrichment of actives among the top-scoring virtual hits, but can anticipate surprising new ideas for synthesis and activity testing. This intended permissiveness ('fuzziness')27 of the CATS molecular representation, which is achieved by coarse-grained atom-typing and feature pair correlation, not only enables scaffold-hopping but may also be used for predicting mutual targets of structurally diverse bioactive ligands. Here, we started from an Ugi-type three-component combinatorial synthesis (Scheme 1)28 and tested whether we could use CATS for 'de-orphanizing' some of the compounds by target identification. All prospective experiments were carried out with the CATS2 implementation. Ugi-type three-component reaction used for constructing a virtual combinatorial library and synthesizing selected compounds. We constructed a virtual combinatorial library from 12 aminopyridines, 40 aldehydes and 8 isocyanide building blocks, resulting in 3840 virtual products (Scheme 1 and Supporting Information). To predict potential bioactivities for these compounds we computed their CATS similarity values to known drugs and lead structures (COBRA v11.10). Briefly, we trained a self-organizing neural network (SOM, Kohonen network) on the pool of COBRA reference compounds and the virtual combinatorial products, followed by visualization of compound distributions as a two-dimensional toroidal map (Figure 4).29,30 For the purpose of prediction, we only considered annotated targets of the reference compounds that were co-clustered with the combinatorial products. In this way, target predictions are limited to a conservative 'application domain' of a reference compound cluster, and the risk of false-positive prediction is reduced.31 For further target prioritization, we computed p-values from the similarity score distribution between ligands binding to different targets (complete training data).32 The p-values are an estimate of the probability of making a false-positive prediction (type-I error). Toroidal self-organizing map (SOM) visualizing overall data density (A), distribution of the virtual combinatorial library (B), known PI3K inhibitors (C), and known muscarinic re

Shown for reading. Its publisher allows quoting it, not republishing it.

References 60

1. 1.

   Pammolli F, Magazzini L, Riccaboni M. Nat. Rev. Drug Discov. 2011;10:428-438.

   Not in our library10.1038/nrd3405

2. 2.

   Paul SM, Mytelka DS, Dunwiddie CT, Persinger CC, Munos BH, Lindborg SR, Schacht AL. Nat. Rev. Drug Discov. 2010;9:203-214.

   Not in our library10.1038/nrd3078

3. 3.

   Schneider G. Nat. Rev. Drug Discov. 2010;9:273-276.

   Not in our library10.1038/nrd3139

4. 4.

   Willett P. Methods Mol. Biol. 2011;672:133-158.

   Not in our library10.1007/978-1-60761-839-3_5

5. 5.

   Vogt M, Bajorath J. Methods Mol. Biol. 2011;672:159-173.

   Not in our library10.1007/978-1-60761-839-3_6

6. 6.

   Gardiner EJ, Holliday JD, O’Dowd C, Willett P. Future Med. Chem. 2011;3:405-414.

   Not in our library10.4155/fmc.11.4

7. 7.

   Renner S, Schneider G. ChemMedChem. 2006;1:181-185.

   Not in our library10.1002/cmdc.200500005

8. 8.

   Garcia-Serna R, Mestres J. Expert Opin. Drug Metab. Toxicol. 2010;6:1253-1263.

   Not in our library10.1517/17425255.2010.509343

9. 9.

   Poroikov V, Filimonov D, Lagunin A, Gloriozova T, Zakharov AB. SAR QSAR Environ. Res. 2007;18:101-110.

   Not in our library10.1080/10629360601054032

10. 10.

    Areias FM, Brea J, Gregori-Puigjané E, Zaki ME, Carvalho MA, Domínguez E, Gutiérrez-de-Terán H, Proença MF, Loza MI, Mestres J. J. Bioorg. Med. Chem. 2010;18:3043-3052.

    Not in our library10.1016/j.bmc.2010.03.048

11. 11.

    Gregori-Puigjané E, Mestres J. Comb. Chem. High Throughput Screen. 2008;11:669-676.

    Not in our library10.2174/138620708785739952

12. 12.

    Gregori-Puigjané E, Mestres J. Curr. Opin. Chem. Biol. 2008;12:359-365.

    Not in our library10.1016/j.cbpa.2008.03.015

13. 13.

    Lounkine E, Keiser MJ, Whitebread S, Mikhailov D, Hamon J, Jenkins JL, Lavan P, Weber E, Doak AK, Côté S, Shoichet BK, Urban L. Nature. 2012;486:361-367.

    Not in our library10.1038/nature11159

14. 14.

    Schneider G, Neidhart W, Giller T, Schmid G. Angew. Chem. Int. Ed. 1999;38:2894-2896.

    Not in our library

15. 15.

    Fechner U, Franke L, Renner S, Schneider P, Schneider G. J. Comput. Aided Mol. Des. 2003;17:687-98.

    Not in our library10.1023/b:jcam.0000017375.61558.ad

16. 16.

    Schneider G, Schneider P, Renner S. QSAR Comb. Sci. 2006;25:1162-1171.

    Not in our library

17. 17.

    Koutsoukas A, Simms B, Kirchmair J, Bond PJ, Whitmore AV, Zimmer S, Young MP, Jenkins JL, Glick M, Glen RC, Bender A. J. Proteomics. 2011;74:2554-2574.

    Not in our library10.1016/j.jprot.2011.05.011

18. 18.

    Achenbach J, Tiikkainen P, Franke L, Proschak E. Future Med. Chem. 2011;3:961-968.

    Not in our library10.4155/fmc.11.62

19. 19.

    Schneider P, Schneider G. QSAR Comb. Sci. 2003;22:713-718.

    Not in our library

20. 20.

    Willett P. Drug Discov. Today. 2006;11:1046-1053.

    Not in our library10.1016/j.drudis.2006.10.005

21. 21.

    Wassermann AM, Bajorath J. Expert Opin. Drug Discov. 2011;6:683-687.

    Not in our library10.1517/17460441.2011.579100

22. 22.

    Kramer C, Kalliokoski T, Gedeck P, Vulpetti A. J. Med. Chem. 2012;55:5165-5173.

    Not in our library10.1021/jm300131x

23. 23.

    Truchon JF, Bayly CI. J. Chem. Inf. Model. 2007;47:488-508.

    Not in our library10.1021/ci600426e

24. 24.

    Bemis GW, Murcko MA. J. Med. Chem. 1996;39:2887-2893.

    Not in our library10.1021/jm9602928

25. 25.

    Rogers D, Hahn M. J. Chem. Inf. Model. 2010;50:742-754.

    Not in our library10.1021/ci100050t

26. 26.

    Carhart RE, Smith DH, Venkataraghavan R. J. Chem. Inf. Comput. Sci. 1985;25:64-73.

    Not in our library

27. 27.

    This reference’s wording can’t be shown under the paper’s licence.

    Not in our library

28. 28.

    This reference’s wording can’t be shown under the paper’s licence.

    Not in our library

29. 29.

    Landrum G.

    Not in our library

30. 30.

    Fechner U, Schneider G. QSAR Comb. Sci. 2004;23:19-22.

    Not in our library

31. 31.

    Holliday JD, Kanoulas E, Malim N, Willett P. J. Cheminform. 2011;3:29.

    Not in our library10.1186/1758-2946-3-29

32. 32.

    Wilcoxon F. Biometrics Bull. 1945;1:80-83.

    Not in our library

33. 33.

    Ward JH. J. Am. Stat. Assoc. 1963;58:236-244.

    Not in our library

34. 34.

    Sastry M, Lowrie JF, Dixon SL, Sherman W. J. Chem. Inf. Model. 2010;50:771-784.

    Not in our library10.1021/ci100062n

35. 35.

    Klenner A, Hartenfeller M, Schneider P, Schneider G. Drug Discov. Today Technol. 2010;7:e237-e244.

    Not in our library

36. 36.

    Schneider G.

    Not in our library10.1016/j.ddtec.2012.06.001

37. 37.

    Bienaymé H, Bouzid K. Angew. Chem. Int. Ed. 1998;37:2234-2237.

    Not in our library10.1002/(sici)1521-3773(19980904)37:16<2234::aid-anie2234>3.0.co;2-r

38. 38.

    Schneider G, Nettekoven M. J. Comb. Chem. 2003;5:233-237.

    Not in our library10.1021/cc020092j

39. 39.

    Schneider P, Tanrikulu Y, Schneider G. Curr. Med. Chem. 2009;16:258-266.

    Not in our library10.2174/092986709787002655

40. 40.

    Schneider P, Stutz K, Kasper L, Haller S, Reutlinger M, Reisen F, Geppert T, Schneider G. Pharmaceuticals. 2011;4:1236-1247.

    Not in our library

41. 41.

    Schneider G, Tanrikulu Y, Schneider P. Future Med. Chem. 2009;1:213-218.

    Not in our library10.4155/fmc.09.11

42. 42.

    Schmuker M, de Bruyne M, Hähnel M, Schneider G. Chem. Cent. J. 2007;1:11.

    Not in our library10.1186/1752-153x-1-11

43. 43.

    Schmuker M, Schneider G. Proc. Natl. Acad. Sci. USA. 2007;104:20285-20289.

    Not in our library10.1073/pnas.0705683104

44. 44.

    Hähnke V, Todoroff N, Rodrigues T, Schneider G. Future Med. Chem. 2012;4:1897-1906.

    Not in our library10.4155/fmc.12.148

45. 45.

    Schneider G. ACS National Meeting, San Diego, CA, USA, March 25–29. 2012.

    Not in our library

46. 46.

    Kim O, Jeong Y, Lee H, Hong SS, Hong S. J. Med. Chem. 2011;54:2455-2466.

    Not in our library10.1021/jm101582z

47. 47.

    Yu HG, Ai YW, Yu LL, Zhou XD, Liu J, Li JH, Xu XM, Liu S, Chen J, Liu JF. Int. J. Cancer. 2008;122:433-443.

    Not in our library10.1002/ijc.23049

48. 48.

    Baviskar AT, Madaan C, Preet R, Mohapatra P, Jain V, Agarwal A, Guchhait SK, Kundu CN, Banerjee UC, Bharatam PV. J. Med. Chem. 2011;54:5013-5030.

    Not in our library10.1021/jm200235u

49. 49.

    Coan KE, Shoichet BK. J. Am. Chem. Soc. 2008;130:9606-9612.

    Not in our library10.1021/ja802977h

50. 50.

    Noeske T, Jirgensons A, Starchenkovs I, Renner S, Jaunzeme I, Trifanova D, Hechenberger M, Bauer T, Kauss V, Parsons CG, Schneider G, Weil T. ChemMedChem. 2007;2:1763-1773.

    Not in our library10.1002/cmdc.200700151

51. 51.

    Renner S, Hechenberger M, Noeske T, Böcker A, Jatzke C, Schmuker M, Parsons CG, Weil T, Schneider G. Angew. Chem. Int. Ed. Engl. 2007;46:5336-5339.

    Not in our library10.1002/anie.200604125

52. 52.

    Steri R, Achenbach J, Steinhilber D, Schubert-Zsilavecz M, Proschak E. Biochem. Pharmacol. 2012;83:1674-1681.

    Not in our library10.1016/j.bcp.2012.02.027

53. 53.

    Reutlinger M, Guba W, Martin RE, Alanine AI, Hoffmann T, Klenner A, Hiss JA, Schneider P, Schneider G. Angew. Chem. Int. Ed. Engl. 2011;50:11633-11636.

    Not in our library10.1002/anie.201105156

54. 54.

    Schneider G, Hartenfeller M, Reutlinger M, Tanrikulu Y, Proschak E, Schneider P. Trends Biotechnol. 2009;27:18-26.

    Not in our library10.1016/j.tibtech.2008.09.005

55. 55.

    Keiser MJ, Roth BL, Armbruster BN, Ernsberger P, Irwin JJ, Shoichet BK. Nat. Biotech. 2007;25:197-206.

    Not in our library10.1038/nbt1284

56. 56.

    Gaulton A, Bellis L, Chambers J, Davies M, Hersey A, Light Y, McGlinchey S, Akhtar R, Atkinson F, Bento AP, Al-Lazikani B, Michalovich D, Overington JP. Nucl. Acids Res. 2012;40:D1100-1107.

    Not in our library10.1093/nar/gkr777

57. 57.

    Sur C, Mallorga PJ, Wittmann M, Jacobson MA, Pascarella D, Williams JB, Brandish PE, Pettibone DJ, Scolnick EM, Conn PJ. Proc. Natl. Acad. Sci. USA. 2003;100:13674-13679.

    Not in our library10.1073/pnas.1835612100

58. 58.

    This reference’s wording can’t be shown under the paper’s licence.

    Not in our library

59. 59.

    Oprea TI, Mestres J. AAPS J. 2012;14:759-763.

    Not in our library10.1208/s12248-012-9390-1

60. 60.

    Jalencas X, Mestres J. Med. Chem. Commun. 2013;4:80-87.

    Not in our library

Cited by papers we hold 2

• Leveraging molecular structure and bioactivity with chemical language models for de novo drug design · 2023
  No notice
• Molecular de-novo design through deep reinforcement learning · 2017
  No notice

Reads

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.