This repository contains code for the paper Predicting Drug Responses of Unseen Cell Types through Transfer Learning with Foundation Models.
Follow the code below to install CRISP. Installation may take about 1 minute.
git clone https://github.com/ml4bio/CRISP.git
cd CRISP
pip install -e .Training:
Follow the tutorial notebook for training. Each training may take 30-60 minutes depends on size of dataset. Or you can directly train with script below. Files of configs and shell scripts are provided in experiments/ for replication of results.
python CRISP/train_script.py --config [path/to/config.yaml] --split split --seed 0 --savedir [path/to/save/folder]Prediction with trained model:
Follow the tutorial notebook. We provide the trained model parameter from Neurips for running prediction.
Preprocessed datasets used in this work all can be downloaded here. There are four perturbation datasets: NeurIPS, SciPlex3, GBM, PC9, and one normal dataset PBMC-Bench. The code of data preprocessing is provided in data folder