LLMsUnderstandMol

This is the code repository for our paper Empirical Evidence for the Fragment level understanding on Drug Molecular Structure of LLMs, which was accepted by AAAI 2024 LLMs4Bio workshop.

Dependencies

pytorch==1.12.1
rdkit==2020.03
tqdm
tensorboard
guacamol

The ChEMBL dataset is available at ChEMBL.

Pre-training

python codes/pretrain.py 

Reinforcement Learning

python codes/generator_guacamol.py --task_id 0
python codes/generator_guacamol.py --task_id 1
python codes/generator_guacamol.py --task_id 2

Analysis of Fragments

See the .ipynb notebooks.