MolP-PC: a multi-view fusion and multi-task learning framework for drug ADMET property prediction

Abstract

The accurate prediction of drug absorption, distribution, metabolism, excretion, and toxicity (ADMET) properties represents a crucial step in early drug development for reducing failure risk. Current deep learning approaches face challenges with data sparsity and information loss due to single-molecule representation limitations and isolated predictive tasks. This research proposes molecular properties prediction with parallel-view and collaborative learning (MolP-PC), a multi-view fusion and multi-task deep learning framework that integrates 1D molecular fingerprints (MFs), 2D molecular graphs, and 3D geometric representations, incorporating an attention-gated fusion mechanism and multi-task adaptive learning strategy for precise ADMET property predictions. Experimental results demonstrate that MolP-PC achieves optimal performance in 27 of 54 tasks, with its multi-task learning (MTL) mechanism significantly enhancing predictive performance on small-scale datasets and surpassing single-task models in 41 of 54 tasks. Additional ablation studies and interpretability analyses confirm the significance of multi-view fusion in capturing multi-dimensional molecular information and enhancing model generalization. A case study examining the anticancer compound Oroxylin A demonstrates MolP-PC’s effective generalization in predicting key pharmacokinetic parameters such as half-life (T0.5) and clearance (CL), indicating its practical utility in drug modeling. However, the model exhibits a tendency to underestimate volume of distribution (VD), indicating potential for improvement in analyzing compounds with high tissue distribution. This study presents an efficient and interpretable approach for ADMET property prediction, establishing a novel framework for molecular optimization and risk assessment in drug development.

Keywords

Molecular ADMET prediction / Multi-view fusion / Attention mechanism / Multi-task deep learning.

References

[1] Pammolli F, Magazzini L, Riccaboni M, et al. The productivity crisis in pharmaceutical R&D. Nat Rev Drug Discov. 2011; 10(6):428-438. https://doi.org/10.1038/nrd3405.

[2] Sun D, Gao W, Hu H, et al. Why 90% of clinical drug development fails and how to improve it? Acta Pharm Sin B. 2022; 12(7):3049-3062. https://doi.org/10.1016/j.apsb.2022.02.002.

[3] Wang Y, Xing J, Xu Y, et al. In silico ADME/T modelling for rational drug design. Q Rev Biophys. 2015; 48(4):488-515. https://doi.org/10.1017/S0033583515000190.

[4] Wei Y, Li S, Li Z, et al. Interpretable-ADMET: a web service for ADMET prediction and optimization based on deep neural representation. Bioinformatics. 2022; 38(10):2863-2871. https://doi.org/10.1093/bioinformatics/btac192.

[5] Jia CY, Li JY, Hao GF, et al. A drug-likeness toolbox facilitates ADMET study in drug discovery. Drug Discov Today. 2020; 25(1):248-258. https://doi.org/10.1016/j.drudis.2019.10.014.

[6] Cherkasov A, Muratov EN, Fourches D, et al. QSAR modeling: where have you been? where are you going to? J Med Chem. 2014; 57(12):4977-5010. https://doi.org/10.1021/jm4004285.

[7] Li Z, Jiang M, Wang S, et al. Deep learning methods for molecular representation and property prediction. Drug Discov Today. 2022; 27(12):103373. https://doi.org/10.1016/j.drudis.2022.103373.

[8] Weininger D. SMILES, a chemical language and information system. 1. Introduction to methodology and encoding rules. J Chem Inf Comput Sci. 1988; 28(1):31-36. https://doi.org/10.1021/ci00057a005.

[9] Weininger D, Weininger A, Weininger JL, et al. SMILES. 2. Algorithm for generation of unique SMILES notation. J Chem Inf Comput Sci. 1989; 29(2): 97-101. https://doi.org/10.1021/ci00062a008.

[10] Durant JL, Leland BA, Henry DR, et al. Reoptimization of MDL keys for use in drug discovery. J Chem Inf Comput Sci. 2002; 42(6):1273-1280. https://doi.org/10.1021/ci010132r.

[11] Cereto-Massagué A, Ojeda MJ, Valls C, et al. Molecular fingerprint similarity search in virtual screening. Methods. 2015; 71:58-63. https://doi.org/10.1016/j.ymeth.2014.08.005.

[12] Steffen A, Kogej T, Tyrchan C, et al. Comparison of molecular fingerprint methods on the basis of biological profile data. J Chem Inf Model. 2009; 49(2):338-347. https://doi.org/10.1021/ci800326z.

[13] Pires DEV, Blundell TL, Ascher DB, et al. pkCSM: predicting small-molecule pharmacokinetic and toxicity properties using graph-based signatures. J Med Chem. 2015; 58(9):4066-4072. https://doi.org/10.1021/acs.jmedchem.5b00104.

[14] Wu CK, Zhang XC, Yang ZJ, et al. Learning to SMILES: BAN-based strategies to improve latent representation learning from molecules. Brief Bioinform. 2021; 22(6):bbab327. https://doi.org/10.1093/bib/bbab327.

[15] Fang Y, Pan X, Shen HB. De novo drug design by iterative multiobjective deep reinforcement learning with graph-based molecular quality assessment. Bioinformatics. 2023; 39(4):btad157. https://doi.org/10.1093/bioinformatics/btad157.

[16] Peng Y, Lin Y, Jing XY, et al. Enhanced graph isomorphism network for molecular ADMET properties prediction. IEEE Access. 2020; 8:168344-168360. https://doi.org/10.1109/ACCESS.2020.3022850.

[17] Xiong Z, Wang D, Liu X, et al. Pushing the boundaries of molecular representation for drug discovery with the graph attention mechanism. J Med Chem. 2020; 63(16):8749-8760. https://doi.org/10.1021/acs.jmedchem.9b00959.

[18] Li C, Wang J, Niu Z, et al. A spatial-temporal gated attention module for molecular property prediction based on molecular geometry. Brief Bioinform. 2021; 22(5):bbab078. https://doi.org/10.1093/bib/bbab078.

[19] Lin X, Dai L, Zhou Y, et al. Comprehensive evaluation of deep and graph learning on drug-drug interactions prediction. Brief Bioinform. 2023; 24(4):bbad235. https://doi.org/10.1093/bib/bbad235.

[20] Zhong S, Zhang K, Wang D, et al. Shedding light on “Black Box” machine learning models for predicting the reactivity of HO radicals toward organic compounds. Chem Eng J. 2021;405:126627. https://doi.org/10.1016/j.cej.2020.126627.

[21] Tang Q, Nie F, Zhao Q, et al. A merged molecular representation deep learning method for blood-brain barrier permeability prediction. Brief Bioinform. 2022; 23(5):bbac357. https://doi.org/10.1093/bib/bbac357.

[22] Karim A, Riahi V, Mishra A, et al. Quantitative toxicity prediction via meta ensembling of multitask deep learning models. ACS Omega. 2021; 6(18):12306-12317. https://doi.org/10.1021/acsomega.1c01247.

[23] Guo Z, Yu W, Zhang C, et al. GraSeq: graph and sequence fusion learning for molecular property prediction. In:Proceedings of the 29th ACM International Conference on Information & Knowledge Management. 2020;435-443. https://doi.org/10.1145/3340531.3411981.

[24] Cai H, Zhang H, Zhao D, et al. FP-GNN: a versatile deep learning architecture for enhanced molecular property prediction. Brief Bioinform. 2022; 23(6):bbac408. https://doi.org/10.1093/bib/bbac408.

[25] Huang Z, Fan Z, Shen S, et al. MolMVC: enhancing molecular representations for drug-related tasks through multi-view contrastive learning. Bioinformatics. 2024; 40(Suppl 2):ii190-ii197. https://doi.org/10.1093/bioinformatics/btae386.

[26] Zhang Y, Yang Q.An overview of multi-task learning. Natl Sci Rev. 2018; 5(1):30-43. https://doi.org/10.1093/nsr/nwx105.

[27] Wu Z, Ramsundar B, Feinberg EN, et al. MoleculeNet: a benchmark for molecular machine learning. Chem Sci. 2018; 9(2):513-530. https://doi.org/10.1039/C7SC02664A.

[28] Huang K, Fu T, Gao W, et al. Therapeutics data commons: machine learning datasets and tasks for drug discovery and development. In: NeurIPS Datasets and Benchmarks Track. 2021. 2021. https://zitniklab.hms.harvard.edu/publications/papers/TDC-neurips21-main.pdf.

[29] Xu Q, Liu K, Lin X, et al. ADMETNet: the knowledge base of pharmacokinetics and toxicology network. J Genet Genomics. 2017; 44(5):273-276. https://doi.org/10.1016/j.jgg.2017.04.005.

[30] Richard AM, Huang R, Waidyanatha S, et al. The Tox 21 10K compound library: collaborative chemistry advancing toxicology. Chem Res Toxicol. 2021; 34(2):189-216. https://doi.org/10.1021/acs.chemrestox.0c00264.

[31] Yang J, Cai Y, Zhao K, et al. Concepts and applications of chemical fingerprint for hit and lead screening. Drug Discov Today. 2022; 27(11):103356. https://doi.org/10.1016/j.drudis.2022.103356.

[32] Stiefl N, Watson IA, Baumann K, et al. ErG: 2D pharmacophore descriptions for scaffold hopping. J Chem Inf Model. 2006; 46(1):208-220. https://doi.org/10.1021/ci050457y.

[33] Shen WX, Zeng X, Zhu F, et al. Out-of-the-box deep learning prediction of pharmaceutical properties by broadly learned knowledge-based molecular representations. Nat Mach Intell. 2021; 3(4):334-343. https://doi.org/10.1038/s42256-021-00301-6.

[34] Bolton EE, Wang Y, Thiessen PA, et al. Chapter 12-PubChem: integrated platform of small molecules and biological activities. Annu Rep Comput Chem. Amsterdam: Elsevier. 2008; 4:217-241. https://doi.org/10.1016/S1574-1400(08)00012-1.

[35] Schlichtkrull M, Kipf TN, Bloem P, et al. Modeling relational data with graph convolutional networks. Lecture Notes in Computer Science. London: Springer Nature. 2018: 593-607. https://doi.org/10.1007/978-3-319-93417-4_38.

[36] Sunseri J, Koes DR. libmolgrid: graphics processing unit accelerated molecular gridding for deep learning applications. J Chem Inf Model. 2020; 60(3): 1079-1084. https://doi.org/10.1021/acs.jcim.9b01145.

[37] Simm J, Humbeck L, Zalewski A, et al. Splitting chemical structure data sets for federated privacy-preserving machine learning. J Cheminform. 2021; 13(1):96. https://doi.org/10.1186/s13321-021-00576-2.

[38] Heid E, Greenman KP, Chung Y, et al. Chemprop: a machine learning package for chemical property prediction. J Chem Inf Model. 2024; 64(1):9-17. https://doi.org/10.1021/acs.jcim.3c01250.

[39] Xiong G, Wu Z, Yi J, et al. ADMETlab 2.0: an integrated online platform for accurate and comprehensive predictions of ADMET properties. Nucleic Acids Res. 2021; 49(W1):W5-W14. https://doi.org/10.1093/nar/gkab255.

[40] Zhao K, Zhou Y, Qiao C, et al. Oroxylin A promotes PTEN-mediated negative regulation of MDM2 transcription via SIRT3-mediated deacetylation to stabilize p53 and inhibit glycolysis in wt-p53 cancer cells. J Hematol Oncol. 2015; 8(1):41. https://doi.org/10.1186/s13045-015-0137-1.

[41] Dai Q, Yin Q, Wei L, et al. Oroxylin A regulates glucose metabolism in response to hypoxic stress with the involvement of Hypoxia-inducible factor-1 in human hepatoma HepG2 cells. Mol Carcinog. 2016; 55(8):1275-1289. https://doi.org/10.1002/mc.22369.

[42] Wei L, Zhou Y, Qiao C, et al. Oroxylin A inhibits glycolysis-dependent proliferation of human breast cancer via promoting SIRT3-mediated SOD2 transcription and HIF1α destabilization. Cell Death Dis. 2015; 6(4):e1714. https://doi.org/10.1038/cddis.2015.86.

[43] Sun X, Chang X, Wang Y, et al. Oroxylin A suppresses the cell proliferation, migration, and EMT via NF- κB signaling pathway in human breast cancer cell. BioMed Res Int. 2019; 2019:1-10. https://doi.org/10.1155/2019/9241769.

[44] Yang X, Zhang F, Wang Y, et al. Oroxylin A inhibits colitis-associated carcinogenesis through modulating the IL-6/STAT3 signaling pathway. Inflamm Bowel Dis. 2013; 19(1):1990-2000. https://doi.org/10.1097/MIB.0b013e318293c5e0.