ADMET property prediction and toxicity profiling of CDK2 inhibitors: a practical cheminformatics tutorial using ADMET-AI 2.0 and RDKit
Abstract
Absorption, Distribution, Metabolism, Excretion, and Toxicity (ADMET) prediction plays a central role in modern computer-aided drug discovery by enabling early evaluation of pharmacokinetic behaviour and safety liabilities. This tutorial presents a reproducible computational workflow for large-scale ADMET profiling using ADMET-AI 2.0, a graph neural network-based prediction platform, applied to 200 cyclin-dependent kinase 2 (CDK2) inhibitors curated from ChEMBL release 35. Implemented within an annotated Jupyter notebook, the workflow integrates cheminformatics preprocessing, physicochemical property analysis, drug-likeness evaluation, structural alert screening, absorption and distribution profiling, CYP450-mediated metabolism prediction, excretion assessment, toxicity evaluation, Tox21 pathway analysis, DrugBank percentile benchmarking, and composite multi-parameter optimisation scoring. A total of 104 ADMET-related outputs were generated per compound, including 52 predictive endpoints and 52 corresponding DrugBank approved-drug percentile rankings. The 200 curated CDK2 inhibitors exhibited 100% Lipinski Ro5 compliance, 38.5% predicted CYP3A4 inhibition, 73.5% predicted hERG liability, and 99.5% predicted DILI risk, with a mean composite ADMET score of 0.46 and a mean traffic light green flag count of 9.5/13. The workflow provides an accessible and FAIR-compliant framework for ADMET analysis, cheminformatics education, and reproducible computational drug discovery.Scientific contributionThis work contributes a structured educational framework for teaching and implementing computational ADMET analysis in modern drug discovery. To be explicit about the nature of this contribution: the novelty is pedagogical and integrative rather than methodological, no new prediction model, algorithm, or endpoint is introduced. The study demonstrates how an existing graph neural network-based prediction model (ADMET-AI) can be integrated with open-source cheminformatics workflows to support interpretable and reproducible pharmacokinetic assessment. Beyond endpoint prediction, the tutorial emphasises practical data handling, workflow automation, compound prioritisation, and transparent result interpretation within a Jupyter notebook environment. The inclusion of DrugBank percentile benchmarking, traffic-light visualisation, and composite scoring strategies provides learners and researchers with practical, worked examples of established multi-parameter decision-making approaches during lead optimisation, rather than new decision frameworks in their own right. By combining methodological clarity with FAIR and reproducible research principles, the workflow serves as a scalable resource for cheminformatics education, computational pharmacology training, and ADMET-guided drug development studies.
Related
- Materials Property Prediction with a Site-Resolved Three-Dimensional Voxel Convolution Neural Network
- Closed-loop Auto Research for Molecular Property Prediction: Discovering and Certifying Generalizable Improvements
- Topology-Guided Multi-Level Graph Learning: A Unified Framework for Molecular Property Prediction
- Beyond conventional sequence-based encoding: a generalized Machine Learning pipeline for oligonucleotide property prediction
- Geometry-Aware Line Graph Transformer Pretraining for Molecular Property Prediction
BibTeX
@article{ogungbemiro2026admet,
title = {{ADMET property prediction and toxicity profiling of CDK2 inhibitors: a practical cheminformatics tutorial using ADMET-AI 2.0 and RDKit}},
author = {Festus O. Ogungbemiro and Jane O. Anebi and Felix O. Okunlola and Love A. Onyekwere and Evna M. M. Pereira},
journal = {Journal of Cheminformatics},
year = {2026},
month = nov,
doi = {10.1186/s13321-026-01289-0},
url = {https://link.springer.com/article/10.1186/s13321-026-01289-0},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - JOUR
TI - ADMET property prediction and toxicity profiling of CDK2 inhibitors: a practical cheminformatics tutorial using ADMET-AI 2.0 and RDKit
AU - Festus O. Ogungbemiro
AU - Jane O. Anebi
AU - Felix O. Okunlola
AU - Love A. Onyekwere
AU - Evna M. M. Pereira
PY - 2026
DA - 2026/11/02
JO - Journal of Cheminformatics
DO - 10.1186/s13321-026-01289-0
UR - https://link.springer.com/article/10.1186/s13321-026-01289-0
AB - Absorption, Distribution, Metabolism, Excretion, and Toxicity (ADMET) prediction plays a central role in modern computer-aided drug discovery by enabling early evaluation of pharmacokinetic behaviour and safety liabilities. This tutorial presents a reproducible computational workflow for large-scale ADMET profiling using ADMET-AI 2.0, a graph neural network-based prediction platform, applied to 200 cyclin-dependent kinase 2 (CDK2) inhibitors curated from ChEMBL release 35. Implemented within an annotated Jupyter notebook, the workflow integrates cheminformatics preprocessing, physicochemical property analysis, drug-likeness evaluation, structural alert screening, absorption and distribution profiling, CYP450-mediated metabolism prediction, excretion assessment, toxicity evaluation, Tox21 pathway analysis, DrugBank percentile benchmarking, and composite multi-parameter optimisation scoring. A total of 104 ADMET-related outputs were generated per compound, including 52 predictive endpoints and 52 corresponding DrugBank approved-drug percentile rankings. The 200 curated CDK2 inhibitors exhibited 100% Lipinski Ro5 compliance, 38.5% predicted CYP3A4 inhibition, 73.5% predicted hERG liability, and 99.5% predicted DILI risk, with a mean composite ADMET score of 0.46 and a mean traffic light green flag count of 9.5/13. The workflow provides an accessible and FAIR-compliant framework for ADMET analysis, cheminformatics education, and reproducible computational drug discovery.Scientific contributionThis work contributes a structured educational framework for teaching and implementing computational ADMET analysis in modern drug discovery. To be explicit about the nature of this contribution: the novelty is pedagogical and integrative rather than methodological, no new prediction model, algorithm, or endpoint is introduced. The study demonstrates how an existing graph neural network-based prediction model (ADMET-AI) can be integrated with open-source cheminformatics workflows to support interpretable and reproducible pharmacokinetic assessment. Beyond endpoint prediction, the tutorial emphasises practical data handling, workflow automation, compound prioritisation, and transparent result interpretation within a Jupyter notebook environment. The inclusion of DrugBank percentile benchmarking, traffic-light visualisation, and composite scoring strategies provides learners and researchers with practical, worked examples of established multi-parameter decision-making approaches during lead optimisation, rather than new decision frameworks in their own right. By combining methodological clarity with FAIR and reproducible research principles, the workflow serves as a scalable resource for cheminformatics education, computational pharmacology training, and ADMET-guided drug development studies.
SN - 1758-2946
ER -