ActStabML

Repository with multi-input graph neural network for modeling metal complexes stability constant

How to setup enviroment

1. Install anaconda/miniconda package manager https://docs.conda.io/projects/miniconda/en/latest/
2. Create new environment conda env create -n actstabml
3. Activate this environment conda activate actstabml
4. Now we will install necessary packages
   

If you have GPU in your machine

conda install pytorch==2.0.0 torchvision==0.15.0 torchaudio==2.0.0 pytorch-cuda=11.8 -c pytorch -c nvidia
conda install pyg -c pyg
conda install dgl -c dglteam/label/cu118
conda install pytorch-lightning rdkit optuna -c conda-forge
conda install pip tqdm loguru pandas 
pip install notebook numba scikit-learn selfies tensorboard dgllife

If you use non-GPU machine

conda install pytorch==2.0.0 torchvision==0.15.0 torchaudio==2.0.0 cpuonly -c pytorch
conda install pyg -c pyg
conda install dgl -c dglteam 
conda install pytorch-lightning rdkit optuna -c conda-forge
conda install pip tqdm loguru pandas
pip install notebook numba scikit-learn selfies tensorboard dgllife

5. To train model on default data

5. python3 train.py

6. To use model for stability constant prediction for benzene

python3 predict.py

To change training and prediction data - modify corresponding scripts