Introduction

Source code for Cross-modal Graph Contrastive Learning with Cellular Images.

Paper on Advanced Science

Environments

MIGA 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

MIGA has been tested on Ubuntu 18.04, with eight GPUs (Nvidia RTX4090). Installation should take no longer than 20 minutes on a modern server.

Please refer to environment.yml

Data

CIL dataset originally consists of 919,265 cellular images collected from 30,616 molecular intervention.

The current CIL data in data includes 50 molecules and 1270 corresponding images, we would release the full version after being accepted.

For pre-training

Check the following scripts: GIN:

    python submit.py --config config/miga/miga_gin.yaml

Graph Transformer:

    python submit.py --config config/miga/miga_graphTrans.yaml

For downstream task (Only support GIN)

Check the following scripts:

    finetune_classification.sh
    finetune_regression.sh
    finetune_clinical.sh

MIGA's pretrained model weights

Model	File Size	Update Date	Download Link
molecular pretrain (GIN)	81MB	Aug 17 2022	[model weights]
molecular pretrain (GraphTransformer)	96MB	Feb 05 2023	[model weights]

MIGA representation

molecule and atoms level representation

import torch
from core.network import MIGA
from dataset import process_data

model = MIGA('graph_transformer', is_eval=True)
model.eval()
checkpoint = torch.load('models/miga_graphtrans_256.pth', map_location='cpu')
model.load_state_dict(checkpoint, strict=False)

smiles = 'c1ccc(cc1)C2=NCC(=O)Nc3c2cc(cc3)[N+](=O)[O]'
data = process_data(smiles, 'graph_transformer')
molecule_embeddings = model.get_graph_embedding(data)

Citation

Please cite the following paper if you use this code in your work.

@article{zheng2022cross,
  title={Cross-Modal Graph Contrastive Learning with Cellular Images},
  author={Zheng, Shuangjia and Rao, Jiahua and Zhang, Jixian and Zhou, Lianyu and Xie, Jiancong and Cohen, Ethan and Lu, Wei and Li, Chengtao and Yang, Yuedong},
  journal={Advanced Science},
  pages={2404845},
  year={2022},
  publisher={Wiley Online Library}
}

prokia/MIGA