Skip to main content

QSAR modeling without descriptors using graph convolutional neural networks: the case of mutagenicity prediction

Abstract

Deep neural networks are effective in learning directly from low-level encoded data without the need of feature extraction. This paper shows how QSAR models can be constructed from 2D molecular graphs without computing chemical descriptors. Two graph convolutional neural network-based models are presented with and without a Bayesian estimation of the prediction uncertainty. The property under investigation is mutagenicity: Models developed here predict the output of the Ames test. These models take the SMILES representation of the molecules as input to produce molecular graphs in terms of adjacency matrices and subsequently use attention mechanisms to weight the role of their subgraphs in producing the output. The results positively compare with current state-of-the-art models. Furthermore, our proposed model interpretation can be enhanced by the automatic extraction of the substructures most important in driving the prediction, as well as by uncertainty estimations.

Graphic abstract

This is a preview of subscription content, access via your institution.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7

References

  1. 1.

    LeCun Y, Bengio Y (1995) Convolutional networks for images, speech, and time series. In Arbib MA (ed) The handbook of brain theory and neural networks, vol. 3361(10)

  2. 2.

    Mayr A, Klambauer G, Unterthiner T, Hochreiter S (2016) DeepTox: toxicity prediction using deep learning. Front Environ Sci 3:80. https://doi.org/10.3389/fenvs.2015.00080

    Article  Google Scholar 

  3. 3.

    Gomez-Bombarelli R, Wei JN, Duvenaud D, Hernandez-Lobato JM, Sanchez-Lengeling B, Sheberla D, Aguilera-Iparraguirre J, Hirzel TD, Adams RP, Aspuru-Guzik A (2018) Automatic chemical design using a data-driven continuous representation of molecules. ACS Cent Sci 4:268–276. https://doi.org/10.1021/acscentsci.7b00572

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  4. 4.

    Putin E, Asadulaev A, Ivanenkov Y, Aladinskiy W, Sanchez-Lengeling B, Aspuru-Guzik A, Zhavoronkov A (2018) Reinforced adversarial neural computer for de novo molecular design. J Chem Inf Model 58:1194–1204. https://doi.org/10.1021/acs.jcim.7b00690

    CAS  Article  PubMed  Google Scholar 

  5. 5.

    Segler MH, Preuss M, Waller MP (2018) Planning chemical syntheses with deep neural networks and symbolic. AI. Nature 555:604. https://doi.org/10.1038/nature25978

    CAS  Article  PubMed  Google Scholar 

  6. 6.

    Zhou Z, Li X, Zare RN (2017) Optimizing chemical reactions with deep reinforcement learning. ACS Cent Sci 3:1337–1344. https://doi.org/10.1021/acscentsci.7b00492

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  7. 7.

    Winkler DA, Le TC (2017) Performance of deep and shallow neural networks, the universal approximation theorem, activity cliffs, and QSAR. Mol Inform 36(1–2):1600118

    Article  Google Scholar 

  8. 8.

    Segler MH, Kogej T, Tyrchan C, Waller MP (2017) Generating focused molecule libraries for drug discovery with recurrent neural networks. ACS Cent Sci 4:120–131. https://doi.org/10.1021/acscentsci.7b00512

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  9. 9.

    Smith JS, Isayev O, Roitberg AE (2017) ANI-1: an extensible neural network potential with DFT accuracy at force field computational cost. Chem Sci 8:3192–3203. https://doi.org/10.1039/C6SC05720A

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  10. 10.

    LeCun Y, Bengio Y, Hinton G (2015) Deep learning. Nature 521:436–444. https://doi.org/10.1038/nature14539

    CAS  Article  Google Scholar 

  11. 11.

    Winter R, Montanari F, Noé F (2019) Clevert D-A (2019) Learning continuous and data-driven molecular descriptors by translating equivalent chemical representations. Chem Sci 10:1692. https://doi.org/10.1039/c8sc04175j

    CAS  Article  PubMed  Google Scholar 

  12. 12.

    Bengio Y, Courville A, Vincent P (2013) Representation learning: a review and new perspectives. IEEE Trans Pattern Anal Mach Intell 35(8):1798–1828. https://doi.org/10.1109/TPAMI.2013.50

    Article  PubMed  Google Scholar 

  13. 13.

    Goh GB, Siegel C, Vishnu A, Hodas N, Baker N (2018) How much chemistry does a deep neural network need to know to make accurate predictions? In 2018 IEEE Winter Conference on Applications of Computer Vision (WACV), p 1340–1349

  14. 14.

    Kearnes S, McCloskey K, Berndl M, Pande V, Riley P (2016) Molecular graph convolutions: moving beyond fingerprints. J Comput Aided Mol Des 30(8):595–608. https://doi.org/10.1007/s10822-016-9938-8

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  15. 15.

    Zhou J, Cui G, Zhang Z, Yang C, Liu Z, Wang L, Li C, Sun M (2019) Graph Neural Networks: a review of methods and applications. AI Open 1. https://doi.org/10.1016/j.aiopen.2021.01.001

    Article  Google Scholar 

  16. 16.

    Roy K (ed) (2017) Advances in QSAR modeling: applications in pharmaceutical chemical food agricultural and environmental sciences. Springer International Publishing, Switzerland

    Google Scholar 

  17. 17.

    Ryu S, Kwon Y, Kim WY (2019) A Bayesian graph convolutional network for reliable prediction of molecular properties with uncertainty quantification. Chem Sci 10:8438–8446. https://doi.org/10.1039/C9SC01992H

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  18. 18.

    Benigni R, Bossa C (2008) Structure alerts for carcinogenicity, and the Salmonella assay system: a novel insight through the chemical relational databases technology. Mutat Res 659(3):248–261. https://doi.org/10.1016/j.mrrev.2008.05.003

    CAS  Article  PubMed  Google Scholar 

  19. 19.

    Gini G, Ferrari T, Cattaneo D, Bakhtyari NG, Manganaro A, Benfenati E (2013) Automatic knowledge extraction from chemical structures: the case of mutagenicity prediction. SAR QSAR Environ Res 24(5):365–383. https://doi.org/10.1080/1062936X.2013.773376

    CAS  Article  PubMed  Google Scholar 

  20. 20.

    Benfenati E, Manganaro A, Gini G (2013) VEGA-QSAR: AI inside a platform for predictive toxicology, Workshop Popularize Artificial Intelligence (PAI) 2013 Torino, http://ceur-ws.org/Vol-1107/

  21. 21.

    Miller EC (1981) Miller J A (1981) Searches for ultimate chemical carcinogens and their reactions with cellular macromolecules. Cancer 47:2327–2345. https://doi.org/10.1002/1097-0142(19810515)47:10%3c2327::aid-cncr2820471003%3e3.0.co;2-z

    CAS  Article  PubMed  Google Scholar 

  22. 22.

    Martin YC, Kofron JL, Traphagen LM (2002) Do structurally similar molecules have similar biological activity? J Med Chem 45(19):4350–4358. https://doi.org/10.1021/jm020155c

    CAS  Article  PubMed  Google Scholar 

  23. 23.

    Hansen K, Mika S, Schroeter T, Sutter A, ter Laak A, Steger-Hartmann T, Heinrich N, Müller K (2009) Benchmark data set for in silico prediction of Ames mutagenicity. J Chem Inf Model 49(9):2077–2081. https://doi.org/10.1021/ci900161g

    CAS  Article  PubMed  Google Scholar 

  24. 24.

    Kazius J, McGuire R, Bursi R (2005) Derivation and validation of toxicophores for mutagenicity prediction. J Med Chem 48:312–320. https://doi.org/10.1021/jm040835a

    CAS  Article  PubMed  Google Scholar 

  25. 25.

    Honma M, Kitazawa A, Cayley A, Williams RV, Barber C, Hanser T, Saiakhov R, Chakravarti S, Myatt GJ, Cross KP, Benfenati E, Raitano G, Mekenyan O, Petkov P, Bossa C, Benigni R, Battistelli CL, Giuliani A, Tcheremenskaia O, Rathman J (2019) Improvement of quantitative structure-activity relationship (QSAR) tools for predicting Ames mutagenicity: outcomes of the Ames/QSAR International Challenge Project. Mutagenesis 34(1):3–16. https://doi.org/10.1093/mutage/gey031

    CAS  Article  PubMed  Google Scholar 

  26. 26.

    Gini G, Katrizky A (Eds.) (1999) Predictive toxicology of chemicals: experiences and impact of AI tools, papers from the AAAI Spring Symposium on Predictive toxicology SS-99-01. AAAI Press, Menlo Park, CA

  27. 27.

    An G (1996) The effects of adding noise during backpropagation training on a generalization performance. Neural Comput 8:643–674. https://doi.org/10.1162/neco.1996.8.3.643

    Article  Google Scholar 

  28. 28.

    Weininger M, Weininger A, Weininger JL (1989) SMILES. 2. Algorithm for generation of unique SMILES notation. J Chem Inf Model 29:97–101. https://doi.org/10.1021/ci00062a008

    CAS  Article  Google Scholar 

  29. 29.

    Wu Z, Pan S, Chen F, Long G, Zhang C, Yu PS (2020) A comprehensive survey on Graph Neural Networks. IEEE Trans Neural Netw Learn Syst. https://doi.org/10.1109/TNNLS.2020.2978386

    Article  PubMed  Google Scholar 

  30. 30.

    Kipf T N, Welling M (2017) Semi-supervised classification with graph convolutional networks. Proceedings International Conference on Learning Representations (ICLR 2017). https://openreview.net/pdf?id=SJU4ayYgl

  31. 31.

    Xiong Z, Wang D, Liu X, Zhong F, Wan X, Li X, Li Z, Luo X, Chen K, Jiang H, Zheng M (2020) Pushing the boundaries of molecular representation for drug discovery with the graph attention mechanism. J Med Chem 63(16):8749–8760. https://doi.org/10.1021/acs.jmedchem.9b00959

    CAS  Article  PubMed  Google Scholar 

  32. 32.

    Mnih V, Heess N, Graves A, Kavukcuoglu K (2014) Recurrent Models of Visual Attention. In Proceedings of NIPS. p 2204–2212

  33. 33.

    Lee JB, Rossi RA, Kim S, Ahmed NK, Koh E (2019) Attention models in graphs: a survey. ACM Trans Knowl Discov Data. https://doi.org/10.1145/3363574

    Article  Google Scholar 

  34. 34.

    Velickovic P, Cucurull G, Casanova A, Romero A, Lio P, Bengio Y (2017) Graph attention networks. In Proceedings ICLR. https://doi.org/10.17863/CAM.48429

  35. 35.

    Gal Y, Ghahramani Z (2016) Dropout as a Bayesian approximation: representing model uncertainty in deep learning. In Proceedings of the 33rd International conference on machine learning, PMLR 48: 1050–1059

  36. 36.

    Gal Y, Hron J (2017) Concrete dropout. In Proceedings 31st International Conference on neural information processing systems, December, p 3584–3593

  37. 37.

    Der Kiureghian A, Ditlevsen O (2009) Aleatory or epistemic? does it matter? Struct Saf 31:105–112. https://doi.org/10.1016/j.strusafe.2008.06.020

    Article  Google Scholar 

  38. 38.

    Kendall A, Gal Y (2017) What uncertainties do we need in Bayesian deep learning for computer vision? Advances in neural information processing systems. 5574– 5584

  39. 39.

    Ames BN (1984) The detection of environmental mutagens and potential. Cancer 53:2030–2040. https://doi.org/10.1002/1097-0142(19840515)53:10%3c2034::aid-cncr2820531005%3e3.0.co;2-s

    Article  Google Scholar 

  40. 40.

    Branco P, Torgo L, Ribeiro RP (2015) A survey of predictive modeling under imbalanced distributions. arXiv:1505.01658v2 [cs.LG]

  41. 41.

    Piegorsch WW, Zeiger E (1991) Measuring intra-assay agreement for the Ames salmonella assay. In: Hotorn L (ed) Statistical methods in toxicology. Springer-Verlag, Berlin, pp 35–41

    Chapter  Google Scholar 

  42. 42.

    Zur RM, Jiang Y, Pesce LL, Drukker K (2009) Noise injection for training artificial neural networks: a comparison with weight decay and early stopping. Med Phys 36(10):4810–4818. https://doi.org/10.1118/1.3213517

    Article  PubMed  PubMed Central  Google Scholar 

  43. 43.

    Polishchuk PG (2017) Interpretation of QSAR models: past, present and future. J Chem Inf Model 57(11):2618–2639. https://doi.org/10.1021/acs.jcim.7b00274

    CAS  Article  PubMed  Google Scholar 

  44. 44.

    Benigni R, Bossa C, Jeliazkova N, Netzeva T, Worth A (2008) The Benigni/Bossa rulebase for mutagenicity and carcinogenicity–a module of Toxtree. JRC Rep 43517 1:6

    Google Scholar 

  45. 45.

    Gini G (2018) QSAR: what else? In: Nicolotti O (ed) Computational toxicology: methods and protocols. Humana Press, New York, NY, pp 79–105

    Chapter  Google Scholar 

  46. 46.

    Benfenati E, Golbamaki A, Raitano G, Roncaglioni A, Manganelli S, Lemke F, Norinder U, Lo Piparo E, Honma M, Manganaro A, Gini G (2018) A large comparison of integrated SAR/QSAR models of the Ames test for mutagenicity. SAR QSAR in Environ Res 29(8):591–611. https://doi.org/10.1080/1062936x.2018.1497702

    CAS  Article  Google Scholar 

  47. 47.

    Gini G, Zanoli F, Gamba A, Raitano G, Benfenati E (2019) Could deep learning in neural networks improve the QSAR models? SAR QSAR in Environ Res 30(9):617–642. https://doi.org/10.1080/1062936X.2019.1650827

    CAS  Article  Google Scholar 

  48. 48.

    Gini G, Zanoli F (2020) Machine learning and deep learning methods in ecotoxicological QSAR modeling. In: Roy K (ed) Ecotoxicological QSARs. Springer Nature, Berlin-Heidelberg

    Google Scholar 

  49. 49.

    Gini G (2020) The QSAR similarity principle in the deep learning era: confirmation or revision? Found Chem 22:383–402. https://doi.org/10.1007/s10698-020-09380-6

    Article  Google Scholar 

  50. 50.

    Honma M (2020) An assessment of mutagenicity of chemical substances by (quantitative) structure–activity relationship. Genes Environ 42:23. https://doi.org/10.1186/s41021-020-00163-1

    Article  PubMed  PubMed Central  Google Scholar 

  51. 51.

    Chakravarti SK, Alla SRM (2019) Descriptor free QSAR modeling using deep learning with long short-term memory neural networks. Front Artif Intell. https://doi.org/10.3389/frai.2019.00017

    Article  PubMed  PubMed Central  Google Scholar 

  52. 52.

    Buckner C, Garson J. (2019) Connectionism. The Stanford Encyclopedia of Philosophy, https://plato.stanford.edu/archives/fall2019/entries/connectionism/

Download references

Author information

Affiliations

Authors

Corresponding author

Correspondence to Giuseppina Gini.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary file1 (DOCX 699 kb)

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Hung, C., Gini, G. QSAR modeling without descriptors using graph convolutional neural networks: the case of mutagenicity prediction. Mol Divers 25, 1283–1299 (2021). https://doi.org/10.1007/s11030-021-10250-2

Download citation

Keywords

  • Toxicity prediction
  • Ames test
  • Deep learning
  • Graph convolutional neural network
  • Bayesian uncertainty
  • Structural alerts