Source code for Geometric-aware deep learning enables discovery of natural product-based liposomes for tumor targeting therapy.
LMR requires anaconda with python 3.7 or later, cudatoolkit=11.1 and below packages
torch 1.7.1+cu110
torch-cluster 1.5.9
torch-geometric 1.6.3
torch-scatter 2.0.7
torch-sparse 0.6.10
torch-spline-conv 1.2.1
torchvision 0.8.2+cu110
LMR has been tested on Ubuntu 18.04, with eight GPUs (Nvidia RTX4090). Installation should take no longer than 20 minutes on a modern server.
GLUT1 dataset originally consists of 14231 molecules collected from CHEMBL and LMR dataset consists of 166 molecules from literature.
Check the following scripts:
pretrain.py
Check the following scripts:
finetune_classification.sh