Related Experiment Video
Updated: Jun 27, 2025
08:21
Curation of Computational Chemical Libraries Demonstrated with Alpha-Amino Acids
Published on: April 13, 2022
2.6K
dMXP: A De Novo Small-Molecule 3D Structure Predictor with Graph Attention Networks
Haopeng Ai1, Deyin Wu1, Huihao Zhou1
1Research Center for Drug Discovery, School of Pharmaceutical Sciences, Sun Yat-Sen University, 132 East Circle at University City, Guangzhou 510006, China.
Journal of Chemical Information and Modeling
|April 25, 2024
Summary
This study introduces dMXP, a novel tool using graph attention networks to predict small molecule 3D structures from crystal data. It achieves high accuracy, with 80% of predicted structures closely matching native poses in receptors.
Area of Science:
- Computational Chemistry
- Drug Design
- Structural Biology
Background:
- Accurate 3D structures of small molecules are vital for structure- and ligand-based drug design.
- Bioactive conformations are essential for lead identification, optimization, and methods like 3D shape similarity searches.
- Crystal structures serve as reliable approximations for bioactive conformers due to energetic proximity to binding poses.
Purpose of the Study:
- To develop a de novo small molecular structure predictor (dMXP).
- To leverage graph attention networks and crystal data for accurate 3D structure generation.
- To improve the reliability of small molecule conformation prediction for drug design.
Main Methods:
- Utilized crystal data from the Cambridge Structural Database (CSD).
- Incorporated molecular electrostatic information from density-functional theory (DFT) calculations.
- Employed graph attention networks with topological and atomic partial charge features to encode molecular graphs.
- Addressed inconsistencies in local substructure contributions to the overall molecular structure.
Main Results:
- Developed the dMXP predictor for generating 3D small molecule structures.
- Achieved high accuracy, with root-mean-square deviations (RMSDs) < 2.0 Å for approximately 80% of predicted structures compared to native binding poses.
- Demonstrated the effectiveness of graph attention mechanisms in handling complex molecular structures.
Conclusions:
- dMXP accurately predicts small molecule 3D structures, closely approximating bioactive conformations.
- The method offers a robust approach for generating high-quality ligand conformations essential for drug discovery.
- This tool can significantly enhance the efficiency and reliability of structure- and ligand-based drug design.
Related Concept Videos
02:06Newman Projections
16.8K
Different notations are used to represent the three-dimensional structure of molecules on two-dimensional surfaces. One of the most commonly used representations is the dash-wedge formula. The dashed wedges, solid wedges, and the plane lines indicate the groups situated behind the plane, coming out of the plane, and in the plane, respectively.
The organic molecules rotate across the single bonds leading to numerous temporary three-dimensional structures of varying energy known as...
The organic molecules rotate across the single bonds leading to numerous temporary three-dimensional structures of varying energy known as...
16.8K
02:00Molecular Models
38.3K
Physical models representing molecular architectures of chemical compounds play essential roles in understanding chemistry. The use of molecular models makes it easier to visualize the structures and shapes of atoms and molecules.
38.3K