Awesome

GARel: Genetic Algorithm based Receptor-ligand interaction generator

Boosts the drug-like features and novelty of the molecules in sampling chemical space

overview of the architecture of GARel

Overview

This repository contains the source of GARel, a software for DL-based de novo drug design.

Requirements

Python == 3.7
pytorch >= 1.1.0
openbabel == 2.4.1
RDKit == 2020.09.5
theano == 1.0.5
vina ==1.2.0 README

if utilizing GPU accelerated model training

CUDA==10.2 & cudnn==7.5

Running GARel

Pretrain Generator

Load sourch dataset (https://github.com/micahwang/RELATION/tree/main/data/zinc) and target dataset ( ./data/sars_cov2_pkis.npz).

python model/run_pretrain.py --epoches 150 --steps 5000 --target sars_cov2 --batchsize 256 --decive 0

Run GARel

python run_gen.py --target sars_cov2