Skip to content
DraconiticResearch
Search Sources API
203,203 studies31 sourcesUpdated 58 min ago
Peer-reviewedCC BYJournal of Chemical Theory and Computation

Reliability of AI Methods in Drug Discovery: Evaluation of Boltz-2 for Structure and Binding Affinity Prediction

  • Shunzhou Wan1
  • Xibei Zhang1
  • Xue Xiao1
  • Peter V. Coveney1
  1. 1University College London , , , ,
Open at sourcePDFDOICite

Abstract

Despite continuing hype about the role of AI in drug discovery, no "AI-discovered drugs" have so far received regulatory approval. Here we assess one of the latest AI-based tools in this domain. Boltz-2, a recently developed biomolecular foundation model, aims to bridge the gap between AI efficiency and physics-based precision through a joint "cofolding" approach. In this study, we provide an extensive evaluation of Boltz-2 using two large-scale data sets: 16780 compounds for 3CLPro and 21702 compounds for TNKS2. We compare Boltz-2 predicted structures with traditional docking and binding affinities with binding free energies derived from the physics-based ESMACS protocol. Structural analysis reveals significant global RMSD variations, indicating that Boltz-2 predicts multiple protein conformations and ligand binding positions rather than a single converged pose. Energetic evaluations exhibit only weak to moderate correlations across the global data sets. Furthermore, a focused analysis of the top 100 compounds yields no significant correlation between the Boltz-2 predictions and the binding free energies from fine-grained ESMACS, alongside frequently observed saturation-state errors in Boltz-2 predicted ligand structures. Our results show that Boltz-2 lacks the energetic resolution required for lead identification. These findings highlight the necessity of employing physics-based methods for the reliability and refinement of AI-derived models.

Related

  1. FLOWR.ROOT – A flow matching-based foundation model for joint multi-purpose structure-aware 3D ligand generation and affinity prediction
    Julian Cremer, Tuan Le, Mohammad Mehdi Ghahremanpour +3
    Nature CommunicationsPeer-reviewedCC BYFull text9 citedMedicinal chemistry & drug discoveryOpenAlex
  2. HierDTA: Hierarchical Structure-Aware Representation Learning for Drug–Target Affinity Prediction
    Huangchao Xu, Baohua Zhang, Qian Liu +2
    ACS OmegaPeer-reviewedCC BY-NC-ND 4.0Full textGeneral chemistryCrossrefPMC
  3. Beyond the Ligand Applicability Domain: Generalization and Local Resolution in Structure-Aware Foundation Models for Affinity Prediction
    Pieter B. Burger, Thomas M. Kaiser, Parker Dryja +7
    ChemRxivPreprintCC BYFull textGeneral chemistryChemRxiv
  4. Accelerating Proton Affinity Prediction with Multi-Fidelity Machine Learning
    Debjyoti Bhattacharya, Yifan Liu, Valentino R. Cooper +1
    ChemRxivPreprintLicence not statedGeneral chemistryChemRxiv
  5. Magnetically recyclable CoFe₂O₄@HKUST-1 nanocatalyst for sustainable carbon–nitrogen coupling, and structure-affinity relationship study of the synthesized products via molecular docking
    Firouz Matloubi Moghaddam, Hassan Fazli, Shaghayegh Sadredini +1
    Scientific ReportsPeer-reviewedCC BY-NC-ND 4.0Full textOrganic chemistryCrossref

Cite

BibTeXRIS
BibTeX
@article{wan2026reliability,
  title = {{Reliability of AI Methods in Drug Discovery: Evaluation of Boltz-2 for Structure and Binding Affinity Prediction}},
  author = {Shunzhou Wan and Xibei Zhang and Xue Xiao and Peter V. Coveney},
  journal = {Journal of Chemical Theory and Computation},
  year = {2026},
  month = jul,
  doi = {10.1021/acs.jctc.6c01334},
  url = {https://doi.org/10.1021/acs.jctc.6c01334},
  note = {CC BY},
}
RIS
TY  - JOUR
TI  - Reliability of AI Methods in Drug Discovery: Evaluation of Boltz-2 for Structure and Binding Affinity Prediction
AU  - Shunzhou Wan
AU  - Xibei Zhang
AU  - Xue Xiao
AU  - Peter V. Coveney
PY  - 2026
DA  - 2026/07/29
JO  - Journal of Chemical Theory and Computation
DO  - 10.1021/acs.jctc.6c01334
UR  - https://doi.org/10.1021/acs.jctc.6c01334
AB  - Despite continuing hype about the role of AI in drug discovery, no "AI-discovered drugs" have so far received regulatory approval. Here we assess one of the latest AI-based tools in this domain. Boltz-2, a recently developed biomolecular foundation model, aims to bridge the gap between AI efficiency and physics-based precision through a joint "cofolding" approach. In this study, we provide an extensive evaluation of Boltz-2 using two large-scale data sets: 16780 compounds for 3CLPro and 21702 compounds for TNKS2. We compare Boltz-2 predicted structures with traditional docking and binding affinities with binding free energies derived from the physics-based ESMACS protocol. Structural analysis reveals significant global RMSD variations, indicating that Boltz-2 predicts multiple protein conformations and ligand binding positions rather than a single converged pose. Energetic evaluations exhibit only weak to moderate correlations across the global data sets. Furthermore, a focused analysis of the top 100 compounds yields no significant correlation between the Boltz-2 predictions and the binding free energies from fine-grained ESMACS, alongside frequently observed saturation-state errors in Boltz-2 predicted ligand structures. Our results show that Boltz-2 lacks the energetic resolution required for lead identification. These findings highlight the necessity of employing physics-based methods for the reliability and refinement of AI-derived models.
SN  - 1549-9618
SN  - 1549-9626
ER  -