Reproducibility, validation, and failure modes across classical and AI-driven molecular docking
- 1Department of Pharmaceutical Sciences, College of Health and Human Sciences, North Dakota State University, Fargo, ND, USA
- 2Department of Pharmaceutical Sciences, College of Health and Human Sciences, North Dakota State University, Fargo, ND, USA. [email protected]
- 3Department of Pharmaceutical Sciences, College of Health and Human Sciences, North Dakota State University, Fargo, ND, USA. [email protected]
Abstract
Molecular docking is indispensable across computer‑aided discovery. However, its conclusions often hinge more on modeling choices than on software brand or nominal score. In this review, we argue that docking should be treated explicitly as conditional modeling whose interpretability depends on structural provenance, ligand‑state definition, search‑space design, and validation under deployment‑relevant conditions. This framework is intended as a practical reference for evaluating docking rigor across both academic and applied workflows. We highlight recurrent failure modes, cross‑target score comparisons under non‑comparable states, over‑read scores as affinities, under‑modeled solvation/flexibility, and uncritical use of predicted structures, and show how AI both exacerbates and mitigates these risks. We then propose best practices for modern validation (self‑docking as necessary but insufficient; cross‑docking, decoys, apo/predicted structures, and out‑of‑distribution tests as essential complements) and offer a concise FAIR reporting checklist enabling reuse and audit. Looking forward, we contend that the most valuable advances are those that improve deployment‑relevant reliability, pose plausibility, enrich screening, and enhance robustness across receptor uncertainty, rather than tool novelty alone. This review reframes docking success from "obtaining a pose and a score" to earning confidence through transparent workflows and evaluation aligned with real use. In this context, the term "AI-driven" does not imply replacing physics-based docking, but rather expanding the workflow landscape in which classical and machine learning approaches coexist. This review, therefore, treats docking as a unified decision framework spanning both paradigms, with emphasis on how validation, generalization, and reproducibility requirements evolve in AI-assisted workflows.
Related
- Docking, Molecular Dynamics, QSAR, and Machine Learning in Alzheimer's Small-Molecule Drug Discovery: Current Utility, Key Limitations, and Reproducibility Challenges
- Exploring the Molecular Mechanism of Ganoderic Acid B for Insomnia Based on Metabolomics, Network Pharmacology, Molecular Docking, and Experimental Validation
- Target-Specific De Novo Drug Design via Fine-Tuned Language Models and Molecular Docking, Molecular Dynamics Simulation Validation
- RaX-DT: An Open-Source Platform for Automated Molecular Docking Workflows and Reproducible Pipeline Validation
- Can Molecular Docking Predict Guest Orientation in β-Cyclodextrin? A Cross-Docking and DFT Refinement Study
BibTeX
@article{kittelson2026reproducibility,
title = {{Reproducibility, validation, and failure modes across classical and AI-driven molecular docking}},
author = {Katiana Simões Kittelson and Allana C. F. Martins and Raquel Possemozer Santos and Gizele Celante and Roberto da Silva Gomes},
journal = {Journal of Computer-Aided Molecular Design},
year = {2026},
month = jun,
doi = {10.1007/s10822-026-00849-8},
url = {https://doi.org/10.1007/s10822-026-00849-8},
note = {CC BY},
}
RIS
TY - JOUR
TI - Reproducibility, validation, and failure modes across classical and AI-driven molecular docking
AU - Katiana Simões Kittelson
AU - Allana C. F. Martins
AU - Raquel Possemozer Santos
AU - Gizele Celante
AU - Roberto da Silva Gomes
PY - 2026
DA - 2026/06/01
JO - Journal of Computer-Aided Molecular Design
DO - 10.1007/s10822-026-00849-8
UR - https://doi.org/10.1007/s10822-026-00849-8
AB - Molecular docking is indispensable across computer‑aided discovery. However, its conclusions often hinge more on modeling choices than on software brand or nominal score. In this review, we argue that docking should be treated explicitly as conditional modeling whose interpretability depends on structural provenance, ligand‑state definition, search‑space design, and validation under deployment‑relevant conditions. This framework is intended as a practical reference for evaluating docking rigor across both academic and applied workflows. We highlight recurrent failure modes, cross‑target score comparisons under non‑comparable states, over‑read scores as affinities, under‑modeled solvation/flexibility, and uncritical use of predicted structures, and show how AI both exacerbates and mitigates these risks. We then propose best practices for modern validation (self‑docking as necessary but insufficient; cross‑docking, decoys, apo/predicted structures, and out‑of‑distribution tests as essential complements) and offer a concise FAIR reporting checklist enabling reuse and audit. Looking forward, we contend that the most valuable advances are those that improve deployment‑relevant reliability, pose plausibility, enrich screening, and enhance robustness across receptor uncertainty, rather than tool novelty alone. This review reframes docking success from "obtaining a pose and a score" to earning confidence through transparent workflows and evaluation aligned with real use. In this context, the term "AI-driven" does not imply replacing physics-based docking, but rather expanding the workflow landscape in which classical and machine learning approaches coexist. This review, therefore, treats docking as a unified decision framework spanning both paradigms, with emphasis on how validation, generalization, and reproducibility requirements evolve in AI-assisted workflows.
SN - 0920-654X
SN - 1573-4951
ER -