Skip to main content
SpringerLink
Log in

Comparing predictive ability of QSAR/QSPR models using 2D and 3D molecular representations

  • Published:
Journal of Computer-Aided Molecular Design Aims and scope Submit manuscript

Abstract

Quantitative structure–activity relationship (QSAR) and quantitative structure–property relationship (QSPR) models predict biological activity and molecular property based on the numerical relationship between chemical structures and activity (property) values. Molecular representations are of importance in QSAR/QSPR analysis. Topological information of molecular structures is usually utilized (2D representations) for this purpose. However, conformational information seems important because molecules are in the three-dimensional space. As a three-dimensional molecular representation applicable to diverse compounds, similarity between a test molecule and a set of reference molecules has been previously proposed. This 3D representation was found to be effective on virtual screening for early enrichment of active compounds. In this study, we introduced the 3D representation into QSAR/QSPR modeling (regression tasks). Furthermore, we investigated relative merits of 3D representations over 2D in terms of the diversity of training data sets. For the prediction task of quantum mechanics-based properties, the 3D representations were superior to 2D. For predicting activity of small molecules against specific biological targets, no consistent trend was observed in the difference of performance using the two types of representations, irrespective of the diversity of training data sets.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Log in via an institution

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2

For each value of QM8 and logD of Lipophilicity, RMSE values for the test data set using SVR models and 2D, 3D descriptors are reported

Fig. 3

For each target, the numbers of training compounds against the diversities of training data set are reported

Fig. 4

For each target macromolecule, RMSE values for the test data set against the diversity (similarity) of the training data sets are reported. Models were constructed using SVR. X-axis is the mean pairwise similarity for compounds in the training data sets

Fig. 5

For each target macromolecule, RMSE values for the test data set using different molecular representations and kernel functions are reported. Shape_kernel + Color_kernel is represented by formula (1). Color*Shape is a hadamard product which use ColorTanimoto and ShapeTanimoto. Shape_kernel*Color_kernel is represented by formula (2)

Fig. 6

In (a), RMSE values for the test data set against training data diversity are reported. Random and Code I represent randomly extracted and Code I reference compounds besides active compounds, respectively. In (b), only individual types of compounds were used without combining active compounds

Fig. 7

NN-distance from a test compound, which could not be predicted, to reference compounds is provided. Both represents compounds that could not be predicted by the two methods being compared. Train_active_cpds_only represents compounds that could not be predicted using only active compounds. Random_only represents compounds that could not be predicted using only randomly selected reference compounds. Code I_only represents compounds that could not be predicted using only Code I reference compounds

Similar content being viewed by others

References

  1. Sippl W, Robaa D (2018) Applied chemoinformatics. Wiley-VCH Verlag GmbH & Co, KGaA, Weinheim

    Google Scholar 

  2. Rodríguez-Pérez R, Miyao T, Jasial S, Vogt M, Bajorath J (2018) Prediction of compound profiling matrices using machine learning. ACS Omega 3:4713–4723

    Article  Google Scholar 

  3. Yuan Q, Wei Z, Guan X, Jiang M, Wang S, Zhang S, Li Z (2019) Toxicity prediction method based on multi-channel convolutional neural network. Molecules 24:3383

    Article  CAS  Google Scholar 

  4. Todeschini R, Consonni V (2009) Molecular descriptors for chemoinformatics. Methods and principles in medicinal chemistry. Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

    Book  Google Scholar 

  5. Kuz’min VE, Polishchuk PG, Artemenko AG, Andronati SA (2011) Interpretation of QSAR models based on random forest methods. Mol Inf 30:593–603

    Article  Google Scholar 

  6. Rodríguez-Pérez R, Bajorath J (2019) Interpretation of compound activity predictions from complex machine learning models using local approximations and shapley values. J Med Chem. https://doi.org/10.1021/acs.jmedchem.9b01101

    Article  PubMed  Google Scholar 

  7. Kearnes S, McCloskey K, Berndl M, Pande V, Riley P (2016) Molecular graph convolutions: moving beyond fingerprints. J Comput Aided Mol Des 30:595–608

    Article  CAS  Google Scholar 

  8. Jo J, Kwak B, Choi HS, Yoon S (2020) The message passing neural networks for chemical property prediction on SMILES. Methods. https://doi.org/10.1016/j.ymeth.2020.05.009

    Article  PubMed  Google Scholar 

  9. Cramer RD, Patterson DE, Bunce JD (1988) Comparative molecular field analysis (CoMFA). 1. Effect of shape on binding of steroids to carrier proteins. J Am Chem Soc 110:5959–5967

    Article  CAS  Google Scholar 

  10. Sato T, Yuki H, Takaya D, Sasaki S, Tanaka A, Honma T (2012) Application of support vector machine to three-dimensional shape-based virtual screening using comprehensive three-dimensional molecular shape overlay with known inhibitors. J Chem Inf Model 52:1015–1026

    Article  CAS  Google Scholar 

  11. Hu B, Kuang ZK, Feng SY, Wang D, He SB, Kong DX, Hu B, Kuang ZK, Feng SY, Wang D et al (2016) Three-dimensional biologically relevant spectrum (BRS-3D): shape similarity profile based on PDB ligands as molecular descriptors. Molecules 21:1554

    Article  Google Scholar 

  12. ROCS version 3.2.2.2; OpenEye Scientific Software Inc, Santa Fe, NM.

  13. Hawkins PCD, Skillman AG, Nicholls A (2007) Comparison of shape-matching and docking as virtual screening tools. J Med Chem 50:74–82

    Article  CAS  Google Scholar 

  14. Miyao T, Jasial S, Bajorath J, Funatsu K (2019) Evaluation of different virtual screening strategies on the basis of compound sets with characteristic core distributions and dissimilarity relationships. J Comput Aided Mol Des 33:729–743

    Article  CAS  Google Scholar 

  15. Hu G, Kuang G, Xiao W, Li W, Liu G, Tang Y (2012) Performance evaluation of 2D fingerprint and 3D shape similarity methods in virtual screening. J Chem Inf Model 52:1103–1113

    Article  CAS  Google Scholar 

  16. Bento AP, Gaulton A, Hersey A, Bellis LJ, Chambers J, Davies M, Kruger FA, Light Y, Mak L, McGlinchey S, Nowotka M, Papadatos G, Santos R, Overington JP (2014) The ChEMBL bioactivity database: an update. Nucleic Acids Res 42:1083–1090

    Article  Google Scholar 

  17. Naveja JJ, Vogt M, Stumpfe D, Medina-Franceo JL, Bajorath J (2019) Systematic extraction of analogue series from large compound collections using a new computational compound-core relationship method. ACS Omega 4:1027–1032

    Article  CAS  Google Scholar 

  18. Wu Z, Ramsundar B, Feinberg EN, Gomes J, Geniesse C, Pappu AS, Leswing K, Pande V (2018) MoleculeNet: a benchmark for molecular machine learning. Chem Sci 9:513

    Article  CAS  Google Scholar 

  19. Ramakrishnan R, Hartmann M, Tapavicza E, Lillienfield OAV (2015) Electronic spectra from TDDFT and machine learning in chemical space. J Chem Phys 143:084111

    Article  Google Scholar 

  20. Experimental in vitro DMPK and physicochemical data on a set of publicly disclosed compounds.

  21. Rogers D, Hahn M (2010) Extended-connectivity fingerprints. J Chem Inf Model 50:742–754

    Article  CAS  Google Scholar 

  22. OEChem TK Version 2.3.0; OpeneEye Scientific Software Inc, Santa, Fe, NM

  23. Molecular Operating Environment (MOE) 2019.01; Chemical Computing Group ULC: 1010 Sherbooke St West Suite #910 Montreal QC Canada H3A 2R7

  24. OEOmega TK Version 2.9.1; OpenEye Scientific Software Inc. Santa Fe, NM

  25. Hornik K (1991) Approximation capabilities of multilayer feed forward networks. Neural Netw 4:251–257

    Article  Google Scholar 

  26. Svetnik V, Liaw A, Tong C, Culberson JC, Sheridan RP, Feuston BP (2003) Random forest: a classification and regression tool for compound classification and QSAR modeling. J Chem Inf Comput Sci 42:1947–1958

    Article  Google Scholar 

  27. Wold S, Sjostrom M, Eriksson L (2001) PLS-regression: a basic tool of chemometrics. Chemometr Intell Lab Syst 58:109–130

    Article  CAS  Google Scholar 

  28. Drucker H, Burges CJC, Kaufman L, Smola AJ, Vapnik V (1996) Support vector regression machines. Neural Inf Process Syst 9:155–161

    Google Scholar 

  29. Pytorch Version 1.5.0

  30. Optuna Version 1.3.0

  31. Nair V, Hinton GE (2010) Rectified linear units improve restricted Boltzmann machines. ICML 10:807–814

    Google Scholar 

  32. Chen CH, Tanaka K, Funatsu K (2018) Random forest approach to QSPR study of fluorescence properties combining quantum chemical descriptors and solvent conditions. J Fluoresc 2:695–706

    Article  Google Scholar 

  33. Pedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O, Blondel M, Prettenhofer P, Weiss R, Dubourg V, Vanderplas J, Passos A, Cournapeau D, Brucher M, Perrot M, Duchesnay E (2011) scikit-learn: machine learning in python. J Mach Learn Res 12:2825–2830

    Google Scholar 

  34. Scipy Version 1.5.0

  35. Irwin JJ, Serling T, Mysinger MM, Bolstad ES, Coleman RG (2012) ZINC: a free tool to discover chemistry for biology. J Chem Inf Model 52:1757–1768

    Article  CAS  Google Scholar 

Download references

Acknowledgements

We thank OpenEye Scientific Software Inc., for providing a free academic license of the OpenEye chemistry toolkits. This work was supported by JSPS KAKENHI Grant Number JP20K19922.

Author information

Authors and Affiliations

  1. Graduate School of Science and Technology, Nara Institute of Science and Technology, 8916-5 Takayama-cho, Ikoma, Nara, 630-0192, Japan

    Akinori Sato, Tomoyuki Miyao & Swarit Jasial

  2. Data Science Center, Nara Institute of Science and Technology, 8916-5 Takayama-cho, Ikoma, Nara, 630-0192, Japan

    Tomoyuki Miyao, Swarit Jasial & Kimito Funatsu

  3. Department of Chemical System Engineering, School of Engineering, The University of Tokyo, 7-3-1 Hongo. Bunkyo-ku, Tokyo, 113-8656, Japan

    Kimito Funatsu

Authors
  1. Akinori Sato
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Tomoyuki Miyao
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Swarit Jasial
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Kimito Funatsu
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Kimito Funatsu.

Additional information

Publisher's Note

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

Electronic supplementary material

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Sato, A., Miyao, T., Jasial, S. et al. Comparing predictive ability of QSAR/QSPR models using 2D and 3D molecular representations. J Comput Aided Mol Des 35, 179–193 (2021). https://doi.org/10.1007/s10822-020-00361-7

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s10822-020-00361-7

Keywords

Search

Navigation