Exploring pocket-aware inhibitors of kinase by generative deep learning, molecular docking, and molecular dynamics simulations
Data
The original data comes from kinase data and is stored in the ./data/generated. sdf file. The training data and test data are located in the DL folder.
Feature
The processing for generating molecular features, pocket features, and model parameter files is handled in the ./DL/make_pretrain_data.py file.
Models
Deep learning experiments were conducted on datasets composed of pocket-aware features. The training script is located in ./pretraining.py, and the trained model is saved in the DL directory.
Molecular Dynamics (MD) Simulations
The interpretability files are located in the MD_datasets folder and include interpretability analyses of the reference and candidate molecule complexes. The Morgan_fingerprinting.py file in the MD folder performs clustering analysis on the candidate molecules, while rg_rmsd.py, RRR.ipynb, and others are used for plotting and analyzing MD simulation trajectories.