Related Experiment Video
Updated: Aug 5, 2026
11:32
A Hydrogen-Deuterium Exchange Mass Spectrometry (HDX-MS) Platform for Investigating Peptide Biosynthetic Enzymes
Published on: May 4, 2020
MInt-HDX: Leveraging Hydrogen-Deuterium Exchange Mass Spectrometry and Machine-Learning to Improve Protein-Ligand
Biorxiv : the Preprint Server for Biology
|July 29, 2026
Summary
This study introduces MInt-HDX, a novel framework combining machine learning and physics-based methods to interpret hydrogen-deuterium exchange mass spectrometry (HDX-MS) data. MInt-HDX enhances protein-ligand docking accuracy, accelerating drug discovery.
Area of Science:
- Structural Biology
- Computational Chemistry
- Biophysics
Background:
- Protein structural dynamics are crucial for biological function and drug discovery.
- Hydrogen-deuterium exchange mass spectrometry (HDX-MS) provides insights into protein dynamics but translating this data to atomic resolution is challenging.
- Current integrative strategies often use HDX-MS data with physics-based models for computational simulations.
Purpose of the Study:
- To develop an integrative framework, MInt-HDX, that combines physics-based and machine learning approaches to interpret HDX-MS data for protein-ligand binding.
- To guide small-molecule ligand docking and improve pose selection using differential HDX-MS signatures.
- To bridge the gap between HDX-MS data and structural modeling for accelerated protein-ligand discovery.
Main Methods:
- Developed MInt-HDX, a hybrid physics-based and machine learning framework utilizing eXtreme Gradient Boosting (XGBoost).
- Trained XGBoost on differential HDX-MS signatures from 11 protein-ligand systems (1032 peptides).
- Integrated XGBoost-predicted residues with 3D clustering and geometric algorithms for docking site generation, followed by HDX-MS-informed filtering and scoring for pose ranking.
Main Results:
- MInt-HDX successfully guided small-molecule ligand docking and pose selection.
- Validated across 3 protein-ligand systems, achieving Ligand-RMSD within 3 Å of crystallographic conformations.
- Demonstrated improved performance compared to common physics-based and machine learning docking methods.
Conclusions:
- MInt-HDX effectively integrates HDX-MS data with computational modeling to enhance protein-ligand structural analysis.
- The hybrid approach accelerates protein-ligand discovery pipelines by improving the accuracy of structural modeling.
- This work highlights the potential of machine learning informed by experimental data and physics-based principles in structural biology.
