Systematic Evaluation of Graph Neural Networks for Ligand-Based Virtual Screening on ChEMBL Datasets
Journal of Chemical Information and Modeling · 2026
DOI: 10.1021/acs.jcim.6c02160 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Chemical Information and Modeling, 2026. Verfasst von Haihan Liu und 6 weiteren Autor:innen. Beteiligte Einrichtung: Department of Medicinal Chemistry, State Key Laboratory of Discovery and Utilization of Functional Components in Traditional Chinese Medicine, Shando…. Hauptthemen laut MEDLINE: Graph Neural Networks, Databases, Chemical. Die Arbeit nennt 43 Literaturangaben. Gefördert von: Key Technology Research and Development Program of Shandong Province, Taishan Scholar Project of Shandong Province, National Science and Technology Major Project.
Abstract (Originalsprache)
The performance of target-specific, ligand-based virtual screening models is strongly influenced by dataset characteristics, including data availability, class imbalance, and evaluation strategies. In this work, we perform a systematic evaluation of graph neural networks (GNNs) using a ChEMBL-derived dataset spanning 5,368 targets and 1.59 million activity records, capturing the long-tailed distributions and target-specific imbalances commonly observed in pharmaceutical data. Through a systematic evaluation of multiple GNN architectures, we identify guidelines for model selection: while the Graph Isomorphism Network (GIN) consistently outperforms others on datasets with >100 samples (a mean ROC–AUC up to 0.94), simpler architectures are more robust under extreme data scarcity. Critically, our comparative analysis of splitting strategies reveals that random sampling yields artificially optimistic performance due to structural overlaps, whereas similarity-aware clustering exposes a substantial generalization gap (AUC drop > 0.3), cautioning against prevailing evaluation practices. We further demonstrate that multi-task learning serves as an effective remedy for small-target instability, providing significant and consistent performance gains. To underscore its translational value, we deploy this comprehensive framework in a virtual screening campaign against Mcl-1, yielding a chemically optimized lead, C4 (Ki = 0.58 μM), with verified cellular efficacy. Our findings highlight the importance of task-aware benchmark design and offer a practical strategy for reliable GNN application in drug discovery.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Substanzen
Autor:innen und Institutionen
- Department of Medicinal Chemistry, State Key Laboratory of Discovery and Utilization of Functional Components in Traditional Chinese Medicine, Shandong Key Laboratory of Druggability Optimization and Evaluation for Lead…
- Phase I Clinical Research Center, Shandong Cancer Hospital, Shandong First Medical University and Shandong Academy of Medical Sciences, Ji'nan, Shandong250117, China.
Förderung
- Key Technology Research and Development Program of Shandong Province
- Taishan Scholar Project of Shandong Province
- National Science and Technology Major Project
- National Natural Science Foundation of China (NSFC)
- Independent Project of Shandong Basic Research Special Zone
Zitieren
APA
Liu, H., Lin, J., Fan, Y., Du, J., Yang, X., Fang, H., & Hou, X. (2026). Systematic Evaluation of Graph Neural Networks for Ligand-Based Virtual Screening on ChEMBL Datasets. Journal of Chemical Information and Modeling. https://doi.org/10.1021/acs.jcim.6c02160
Vancouver
Liu H, Lin J, Fan Y, Du J, Yang X, Fang H, et al. Systematic Evaluation of Graph Neural Networks for Ligand-Based Virtual Screening on ChEMBL Datasets. Journal of Chemical Information and Modeling. 2026. doi:10.1021/acs.jcim.6c02160
BibTeX
@article{Liu2026,
author = {Liu, Haihan and Lin, Jiaqi and Fan, Ying and Du, Jintong and Yang, Xinying and Fang, Hao and Hou, Xuben},
title = {Systematic Evaluation of Graph Neural Networks for Ligand-Based Virtual Screening on ChEMBL Datasets},
journal = {Journal of Chemical Information and Modeling},
year = {2026},
doi = {10.1021/acs.jcim.6c02160}
}
EUVIMED