This survey paper comprehensively explores the landscape of predictive modeling for Absorption, Distribution,
Metabolism, Excretion, and Toxicity (ADMET) properties of drugs through the lens of machine learning (ML) techniques. The
review encompasses an extensive analysis of methodologies, data sets, advancements in ML algorithms, and their applications in
drug discovery and development. Beginning with an overview of the significance of ADMET properties in drug development, the
survey delves into various datasets utilized for modeling, encompassing chemical descriptors, biological activities,
physicochemical properties, and toxicity endpoints. It scrutinizes the intricacies of feature engineering, emphasizing the
importance of selecting informative features for accurate predictions. The survey critically evaluates an array of ML algorithms
employed in predictive modeling, ranging from traditional methods to state-of-the-art deep learning architectures. It highlights the
strengths, limitations, and applications of these algorithms in predicting ADMET properties, emphasizing the need for robust
experimental design and validation protocols. Challenges such as interpretability, data quality, and integration of domain
knowledge are addressed, underscoring the significance of standardized frameworks for ensuring reproducibility and generalizing
ability of predictive models. Furthermore, the survey showcases successful applications of ML-based predictive modeling in
optimizing drug candidate selection, mitigating toxicity risks, and expediting the drug discovery process.
[1] Cheng, F., et al. “Prediction of drug-target interactions and drug repositioning via network-based inference”. PLOS
Computational Biology, 8(5), e1002503, 2012.
[2] Chen, H., et al. “Deep learning in label-free cell classification”, Genome Biology, 2018, 19(1), pp. 1-10.
[3] Mayr, A., et al. “ DeepTox: Toxicity prediction using deep learning” Frontiers in Environmental Science, .2016, pp.80-85.
[4] Xu, Y., et al. “Deep learning for drug- induced liver injury” Journal of Chemical Information and Modeling, 2018 58(3), pp.
487-492.
[5] Wallach, I., & Heifets, A. “Mostlig and-based classification benchmarks reward memorization rather than generalization”
Journal of Chemical Information and Modeling, 2018, 58(5), 916-932.
[6] Yang, K.,& Swanson, K.“Comparative analyses of deep learning and machine learning models for prediction of chemical
toxicity” Journal of Chemical Information and Modeling, 2018, 58(8), 1553-1562.
[7] Zhang, L.,Tan,J.,Han, D.,Zhu,H., & Fromm, M. F. (2016). “ Prediction of drug-drug interactions through drug structural
similarities and interaction networks incorporating pharmaco kinetics and pharmaco dynamics knowledge” Journal of Chemical Information and Modeling, 56(12), 2446-2456.
[8] Ekins, S., et al. (2015). “Progress in predicting human ADME parameters in silico” Journal of Pharmacological and
Toxicological Methods, 71, 126-145.
[9] Fourches, D., et al. “Chem informatics analysis of assertions mined from the literature that describe the drug-induced liver
injury in different species” Chemical Research in Toxicology, 2010, 23(1), 171-183.
[10] Hughes, J. D., et al. (2008). “Principles of early drug discovery”, British Journal of Pharmacology, 152(1), 38-44.
[11] https://towards data science.com/ workflow-of-a-machine-learning-project-ec1dba419b94
[12] Predictive Multitask Deep Neural Network Models for ADME-Tox Properties: Learning from Large Data
Sets]https://pubs.acs.org/doi/abs/10.1021/acs.jcim.8b00785
[13] https://www.mdpi.com/1420-3049/27/9/3021
[14] https://wires.onlinelibrary.wiley.com/doi/epdf/10.1002/wcms.1568
[15] https://www.sciencedirect.com/science/article/pii/S2001037021003421
[16] https://pubs.acs.org/doi/10.1021/acsomega.1c05203
[17] https://pubs.acs.org/doi/abs/10.1021/acs.jcim.3c00160
