TrustworthinessTransparencyExplainabilitycs.LG
Trustworthy Protein-Ligand Binding Affinity Prediction via Reliability-Aware Multi-Engine Fusion
Yongchan Hong, Defu Cao, Wenjin Liu, Thomas Ku, Jordy Homing Lam, Emily Nguyen, Willie Neiswanger, Vsevolod Katritch, Yan Liu
July 20, 2026
Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2
1 citations
Key Takeaways
-
Uncertainty quantification improves drug discovery reliability: The RELIABLE-BA framework provides interpretable confidence measures alongside binding affinity predictions, enabling researchers to identify which predictions to trust—a critical gap since traditional ensemble methods treat all predictions equally without indicating confidence levels.
-
Selective prediction reduces errors significantly: By filtering out low-confidence predictions, the model achieved up to 25% error reduction on retained protein-ligand pairs, demonstrating that knowing when not to trust a prediction is as valuable as making accurate predictions for drug discovery workflows.
-
Chemical context matters for prediction confidence: Unlike standard consensus approaches, RELIABLE-BA learns reliability from molecular context (the specific chemical properties of each protein-ligand pair), meaning confidence scores are tailored to individual cases rather than applying uniform uncertainty estimates across all predictions.
-
Validated on clinically relevant targets: Testing on SARS-CoV-2 Mpro and 5HT2A receptor datasets demonstrates the method's applicability beyond benchmark datasets to real drug discovery scenarios, suggesting practical utility for pharmaceutical research.
-
Addresses a fundamental disagreement problem: Modern docking engines frequently disagree on binding affinity predictions without explanation; this work provides a principled framework for understanding and reconciling these disagreements through evidential reasoning, improving transparency in computational drug discovery.
Abstract
Accurate protein-ligand binding affinity prediction is central to computational drug discovery, yet modern docking engines frequently disagree without indicating which prediction to trust. Consensus scoring and ensemble methods improve mean accuracy but treat all predictions identically without interpretable confidence measures or uncertainty decomposition, ignoring the chemical context of each protein-ligand pair. To address this limitation, we introduce RELIABLE-BA (RELIABiLity-aware Evidential fusion for Binding Affinity), an evidential framework for multi-engine binding affinity prediction. Our model comprises three steps: (1) modeling each engine as an evidential expert via Normal-Inverse-Gamma distributions, (2) scaling epistemic uncertainty through learned reliability from molecular context while preserving each expert's predictive mean, and (3) fusing experts through closed-form aggregation that captures both individual uncertainty and inter-engine disagreement. Experiments on the PDBBind and BDB2020+ benchmarks demonstrate competitive point prediction with substantially improved uncertainty calibration, and additional validation on the SARS-CoV-2 Mpro dataset and 5HT2A receptor demonstrates applicability to clinically relevant drug targets. Crucially, these uncertainty estimates enable reliable filtering of protein-ligand pairs, reducing prediction error by up to 25% when retaining only high-confidence pairs. To our knowledge, RELIABLE-BA is the first multi-engine binding affinity prediction framework to combine evidential fusion with context-dependent reliability, offering a principled path toward trustworthy AI-guided drug discovery. Our code is publicly available at https://github.com/yongchand/RELIABLE-BA.
Paper Info
arXiv ID2607.17601
PublishedJuly 20, 2026
Citations
1
Relevance72%
VenueProceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2
TypeBook, JournalArticle, Conference
Categorycs.LG
Responsible AI Topics
TrustworthinessTransparencyExplainability
Links
Citation data from Semantic Scholar