InteractionGraphNet: A Novel and Efficient Deep Graph Representation Learning Framework for Accurate Protein–Ligand Interaction Predictions
Problem
Predicting how strongly small molecules bind to target proteins is a key bottleneck in structure-based drug design, and traditional ML descriptors struggle with 3D molecular interactions.
Why it matters
Binding prediction guides drug development, affecting the cost and speed of discovering new medicines.
What the AI did
InteractionGraphNet (IGN) uses two stacked graph convolution modules on 3D protein-ligand complex structures to learn intramolecular and intermolecular interactions for prediction tasks.
Results
Better or competitive performance vs. state-of-the-art methods in binding affinity prediction, virtual screening, and pose prediction; no numbers given in abstract.
Impact
Provides a benchmarked deep graph learning tool for binding prediction and virtual screening.
Record
| Venue | Journal of Medicinal Chemistry |
|---|---|
| Published | 2021 |
| Work type | article |
| Authors | Dejun Jiang, Chang‐Yu Hsieh, Zhenxing Wu, Yu Kang, Jike Wang, Ercheng Wang, Ben Liao, Chao Shen et al. |
| Citations | 261 (OpenAlex, at import time) |
| References | 62 |
| Model or method | InteractionGraphNet (IGN) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1021/acs.jmedchem.1c01830 |
| Open access | closed |
| Keywords | protein-ligand interactions, graph neural networks, binding affinity, virtual screening, drug design |
Importance rating
significant — thousands to millions affected
Abstract calls accurate protein-ligand interaction quantification a key challenge to structure-based drug design; no scale numbers given.