Integrating Conformational Sampling and Siamese Learning to Predict Mutation-Induced Binding Affinity Changes in
Liangxu Xie1, Teng Ma2, Xiaohua Lu1
1Institute of Bioinformatics and Medical Engineering, Jiangsu University of Technology, Changzhou 213001, China.
Journal of Chemical Theory and Computation
|June 1, 2026
Summary
Predicting mutation effects on protein-ligand binding affinity is vital for drug discovery. Our new method uses AlphaFold 2 (AF2) subsampling and a Siamese network to accurately model these changes, improving drug development.
Area of Science:
- Computational Biology
- Structural Biology
- Drug Discovery
Background:
- Predicting mutation impacts on protein-ligand binding affinity is critical for drug discovery, addressing resistance and repurposing. Traditional methods struggle with static structures.
- Protein flexibility is key, but often overlooked by conventional structure-based approaches.
Purpose of the Study:
- To develop a novel computational framework for predicting mutation-induced changes in protein-ligand binding affinity.
- To overcome limitations of static structures by incorporating conformational flexibility.
Main Methods:
- Integrated AlphaFold 2 (AF2) subsampling to generate conformational ensembles for protein mutants (e.g., Abelson tyrosine kinase - ABL).
- Developed SIGMA-Net (Siamese structure and graph-aware multistructural affinity prediction network) using a Siamese neural network architecture.
- Augmented data with reference states and identified relevant conformations via most probable distribution analysis.
Main Results:
- The SIGMA-Net approach, leveraging AF2 subsampling, achieved higher correlation coefficients for five of six tyrosine kinase inhibitors (TKI) across 31 ABL mutants compared to docking and TriG-Net.
- Demonstrated robust prediction of relative binding free energy (RBFE) and comparable absolute binding free energy (ABFE) performance to state-of-the-art models.
- The ensemble-based approach significantly improved predictive accuracy and robustness over static structure methods.
Conclusions:
- The proposed framework effectively models protein-ligand interactions in flexible targets by integrating conformational sampling with Siamese learning.
- This method offers a transferable solution, transcending static structure limitations for enhanced drug discovery and development.
- The study highlights the importance of accounting for protein dynamics in predicting binding affinity changes.
Related Concept Videos
01:49Conserved Binding Sites
Many proteins’ biological role depends on their interactions with their ligands, small molecules that bind to specific locations on the protein known as ligand-binding sites. Ligand-binding sites are often conserved among homologous proteins as these sites are critical for protein function.
Binding sites are often located in large pockets, and if their location on a protein’s surface is unknown, it can be predicted using various approaches. The energetic method computationally analyses the...
Binding sites are often located in large pockets, and if their location on a protein’s surface is unknown, it can be predicted using various approaches. The energetic method computationally analyses the...
00:49Ligand Binding and Linkage
Allosteric proteins have more than one ligand binding site; the binding of a ligand to any of these sites influences the binding of ligands to the other sites. When a protein is allosteric, its binding sites are called coupled or linked. In the case of enzymes, the site that binds to the substrate is known as the active site and the other site is known as the regulatory site. When a ligand binds to the regulatory site, this leads to conformational changes in the protein that can influence the...