Linking machine learning and biophysical structural features in drug discovery - Original research

Aug 5, 2026

Upload by Levi

The study links machine learning with biophysical structural features to improve drug discovery by identifying pharmacophore characteristics tied to ligand-selected protein conformations. Molecular dynamics simulations generate an ensemble of protein binding-site conformations capturing binding-site dynamics, while pharmacophore descriptors drive an AI/ML prioritization framework for features unique to ligand-selected states. The method enriches true positive ligands by up to 54-fold versus random selection and is robust across diverse proteins. Unlike conventional screening, the work emphasizes conformational selection, making binding interactions interpretable and actionable for lead optimization.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages10

Reading Time25 min

AccessFree to download

Introduction

Conformational selection and ensemble docking

Machine learning for explainable feature identification

Methods

Molecular dynamics ensemble generation

Pharmacophore descriptors and AI/ML prioritization

Results

Database enrichment performance

Conclusion

Interpretable, mechanism-driven screening

TYPE Original Research PUBLISHED 23 January 2025 DOI 10.3389/fmolb.2024.1305272 OPEN ACCESS EDITED BY Ognjen Perisic, Big Blue Genomics/Redesign Science, Serbia REVIEWED BY Giulia Morra, National Research Council (CNR), Italy Karan Kapoor, Ensem Therapeutics, United States *CORRESPONDENCE Jerome Baudry, [jerome.baudry@uah.edu](jerome.baudry@uah.edu)[ ](jerome.baudry@uah.edu)Vineetha Menon, [vineetha.menon@uah.edu](vineetha.menon@uah.edu) †These authors have contributed equally to this work RECEIVED 01 October 2023 ACCEPTED 27 December 2024 PUBLISHED 23 January 2025 CITATION Ahmadi A, Gupta S, Menon V and Baudry J (2025) Linking machine learning and biophysical structural features in drug discovery. Front. Mol. Biosci. 11:1305272 . doi: 10.3389/fmolb.2024.1305272 COPYRIGHT © 2025 Ahmadi, Gupta, Menon and Baudry. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY) . The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms. Linking machine learning and biophysical structural features in drug discovery Armin Ahmadi 1†, Shivangi Gupta 2†, Vineetha Menon 2* and Jerome Baudry 1* 1 Department of Biological Sciences, The University of Alabama in Huntsville, Huntsville, AL, United States, 2 Department of Computer Science, The University of Alabama in Huntsville, Huntsville, AL, United States Introduction: Machine learning methods were applied to analyze pharmacophore features derived from four protein-binding sites, aiming to identify key features associated with ligand-specific protein conformations. Methods: Using molecular dynamics simulations, we generated an ensemble of protein conformations to capture the dynamic nature of their binding sites. By leveraging pharmacophore descriptors, the AI/ML framework prioritized features uniquely associated with ligand-selected conformations, enabling a mechanism-driven understanding of binding interactions. This novel approach integrates biophysical insights with machine learning, focusing on pharmacophoric properties such as charge, hydrogen bonding, hydrophobicity, and aromaticity. Results: Results showed significant enrichment of true positive ligands—improving database enrichment by up to 54-fold compared to random selection—demonstrating the robustness of this approach across diverse proteins. Conclusion: Unlike conventional structure-based or ligand-based screening methods, this work emphasizes the role of specific protein conformations in driving ligand binding, making the process highly interpretable and actionable for drug discovery. The key innovation lies in identifying pharmacophore features tied to conformations selected by ligands, offering a predictive framework for optimizing drug candidates. This study illustrates the potential of combining MLand pharmacophoric analysis to develop intuitive and mechanism-driven tools for lead optimization and rational drug design. KEYWORDS drug discovery, machine learning, pharmacophore, conformational selection, docking, ensemble docking, chemical biology 1 Introduction Contemporary approaches in computational drug discovery and computational chemical biology are mostly centered around protein: ligand predictions and rationalizations, in particular through using docking approaches. Docking is not only used to identify a specific binding mode for a particular ligand in a given protein target, but it is also used, and maybe mostly so nowadays, as a virtual screening tool that will “reduce the size of the chemical haystack” and allow a faster Frontiers in Molecular Biosciences 01 [frontiersin.org](frontiersin.org) and more economical identif

FAQs

How does the approach generate protein conformational information for screening?

It uses molecular dynamics simulations to generate an ensemble of protein binding-site conformations that reflect the dynamic nature of binding sites, which are then evaluated through docking and feature analysis.

What does the machine learning framework prioritize?

The framework prioritizes pharmacophore features uniquely associated with ligand-selected protein conformations, using pharmacophore descriptors that reflect biophysical properties such as charge, hydrogen bonding, hydrophobicity, and aromaticity.

How much does the method improve enrichment compared with random selection?

The results show significant enrichment of true positive ligands, improving database enrichment by up to 54-fold compared with random selection.

Aug 5, 2026

Upload by Levi