A Reproducible Hierarchical Virtual Screening Framework Integrating Scaffold-Aware Machine Learning, Ensemble Docking, and Molecular Dynamics: Application to IDO1

Clinical Snapshot

Key Findings

• P Value: Not reported in abstract

• Effect Size: ROC-AUC 0.88–0.89 for ML ensemble; specific enrichment factor values not reported in abstract

• Primary Outcome: Ensemble ML classifier (Random Forest, XGBoost, SVM) achieved ROC-AUC ≈0.88–0.89 on scaffold-split validation; best-Z-score consensus docking outperformed mean, rank-based, and weighted aggregation on enrichment factor metrics; 39 FDA-approved compounds predicted active within applicability domain

• Nnt Or Sensitivity: Not applicable (computational preclinical study); sensitivity/specificity values not reported in abstract — ROC-AUC serves as the primary discriminative metric

• Confidence Interval: Not explicitly reported in abstract; nested cross-validation implies internal variance estimation

Clinical Application

The computational workflow is designed for reproducibility and accessibility, with full open-source implementation. Adoption requires computational chemistry expertise, access to molecular dynamics simulation infrastructure, and familiarity with cheminformatics tools. The Jupyter notebook format lowers the barrier to replication. Experimental follow-up (biochemical assays, cellular models) would require standard medicinal chemistry laboratory infrastructure. IDO1 inhibition remains an active area of oncology research in Australia, with interest from groups at institutions including the Peter MacCallum Cancer Centre and QIMR Berghofer. The TGA has not approved any IDO1 inhibitor; epacadostat (Incyte) failed Phase III (ECHO-301/KEYNOTE-252) in combination with pembrolizumab, raising questions about the clinical utility of IDO1 inhibition as a monotherapy or combination strategy. The drug repurposing focus on FDA-approved compounds is relevant to the Australian context given PBS listing considerations — repurposed agents with existing safety profiles face a lower regulatory burden. RACGP guidelines do not currently address IDO1 inhibitors. Australian computational chemistry groups could adopt this open-source framework directly. Any candidates identified would require progression through standard TGA preclinical and clinical development pathways. Not directly applicable to any patient population at this stage. The framework is relevant to medicinal chemists and computational drug discovery scientists working on IDO1-targeted oncology programmes. Potential future relevance to patients with IDO1-overexpressing tumours (e.g., certain colorectal, endometrial, and haematological malignancies) if candidates progress to experimental validation.

Abstract

Indoleamine 2,3-dioxygenase 1 (IDO1) is a heme-containing enzyme implicated in cancer immune escape and remains an attractive therapeutic target despite recent clinical setbacks. We report a fully reproducible hierarchical virtual screening framework integrating scaffold-aware machine learning, ensemble docking, consensus scoring, and molecular dynamics simulations for robust prioritization of IDO1 inhibitors. A curated ChEMBL data set of IDO1 inhibitors was subjected to strict standardization, duplicate removal, and activity binarization at pChEMBL ≥6. Models were trained using scaffold-based splitting and nested cross-validation to prevent chemical series leakage. An ensemble of Random Forest, XGBoost and SVM classifiers achieved balanced predictive performance (ROC-AUC ≈0.88-0.89) with applicability domain filtering to ensure reliability. Prospective screening of FDA-approved drugs yielded 39 compounds within the applicability domain predicted as active. These were further evaluated through ensemble docking against multiple IDO1 crystal structures using GNINA with CNN rescoring. Consensus strategies were systematically benchmarked, demonstrating that best-Z-score aggregation outperformed mean, rank-based, and weighted methods in enrichment factor (EF) metrics. Two top-ranked candidates were subjected to 300 ns molecular dynamics simulations, revealing stable binding modes and persistent interactions with key catalytic residues. This study demonstrates that hierarchical integration of scaffold-aware machine learning with structure-based ensemble strategies enhances robustness and reduces false positives in virtual screening campaigns. The proposed workflow is generalizable and supports reproducible candidate prioritization in computational drug discovery. The complete implementation, including data processing, model training, and analysis steps, is provided as a fully executable Jupyter notebook available at https://github.com/rocco-b/IDO1-inhibitors-ML-and-docking-data.

References

1. 1.Tomarchio, E. G., Buccheri, R., & Rescifina, A. (2026). A reproducible hierarchical virtual screening framework integrating scaffold-aware machine learning, ensemble docking, and molecular dynamics: Application to IDO1. Journal of Chemical Information and Modeling. https://doi.org/10.1021/acs.jcim.6c00967