Xu, Mme Kexin (2023) Incorporating Solvent Information in Graph Diffusion Models for Protein Docking PRE - Research Project, ENSTA.
![]()
| PDF 241Kb |
Abstract
Ligand-protein docking is a crucial process in drug discovery, enabling the exploration of potential binding interactions between small molecule ligands and protein targets. Solvent media plays an essential role in mediating bonds. However, existing docking methods often neglect the explicit treatment of the solvent in which the protein and ligand emerge. In this study, we extend DiffDock, a state-of-the-art docking approach, to efficiently process any type of solvent data, capturing their pivotal roles in ligand binding. We construct our model, DiffDock-Sol, which composes of a graph diffusion model of ligand, protein as well as the solvent for predicting the correct docking pose of ligand, and a confidence model for giving the docking rank. We perform extensive experimental validations based on data from water molecules to assess their predictive power and efficiency. Through experimental validation, we demonstrate the effective predictive power of our docking model as well as the significance of solvent media in docking.
Item Type: | Thesis (PRE - Research Project) |
---|---|
Uncontrolled Keywords: | Ligand-Protein Docking, Solvent, Water Molecule, Graph Diffusion Model, Graph Neural Network |
Subjects: | Information and Communication Sciences and Technologies |
ID Code: | 9577 |
Deposited By: | Kexin XU |
Deposited On: | 24 août 2023 15:20 |
Dernière modification: | 28 août 2023 10:30 |
Repository Staff Only: item control page