Pisces: A multi-modal data augmentation approach for drug combination synergy prediction

This repository is the official implementation of Pisces: A multi-modal data augmentation approach for drug combination synergy prediction.

Requirements and Installation

PyTorch version == 1.8.0
PyTorch Geometric version == 1.6.3
RDKit version == 2020.09.5

You can build the Dockerfile or use the docker image teslazhu/pretrainmol36:latest.

To install the code from source

git clone https://github.com/linjc16/Pisces.git

pip install fairseq
pip uninstall -y fairseq 

pip install ninja
python setup.py build_ext --inplace

Getting Started

Experiments folder

Here we reproduced all three tasks across different settings in our code base, including cell-line-based drug synergy prediction, xenograft-based drug synergy prediction and drug-drug interaction prediction. The mapping between our scripts folder and the particular experiments is as following:

Pisces/  
├── scripts/  
│   ├── gdsc_trans/ # Vanilla cross validation on GDSC-Combo
│   ├── gdsc_leave_comb/ # Split by combination on GDSC-Combo
│   ├── gdsc_leave_cell/ # Split by cell line on GDSC-Combo
│   ├── xenograft_best_response/ # BestResponse prediction on Xenografts
│   ├── xenograft_days_response/ # Drug combination response prediction across all time points on Xenografts
│   ├── xenograft_extrapolation/ # Response prediction at the last time point on Xenografts
│   ├── drugbank_trans/ # Drug-drug interaction vanilla cross validation on DrugBank
│   ├── drugbank_ind/ # One new drug in each test pair on DrugBank
│   ├── drugbank_unseen/ # Two new drugs in each test pair on DrugBank
│   └── two_sides # Vanilla cross validation on TwoSIDES
└── src/

All the experiments follow the similar pipelines to reproduce the results. Now we take vanilla cross validation on GDSC-Combo as an example to illustrate such a process.

Data Preprocessing

We evaluate our models on the dataset above. To generate the binary data for fairseq, take the 5-fold CV setting (fold 0) as an example, run

python Pisces/scripts/gdsc_trans/data_process/run_process.py

bash Pisces/scripts/gdsc_trans/data_process/run_binarize.sh 0

Note that you need to change the file paths accordingly. More original data can be found here.

Training and Test

All training and test scripts can be seen in Pisces/scripts/gdsc_trans. For instance,

bash Pisces/scripts/gdsc_trans/run.sh 0 0

bash Pisces/scripts/train_trans/inference/inf.sh 0 0

Contact

Please feel free to submit a Github issue if you have any questions or find any bugs. We do not guarantee any support, but will do our best if we can help.