MPDP：Multitask-Prediction-between-Drug-Protein
python version 3.7.x
pip install -r requirements.txt
embedding data is already generated just
python mpdp.py
- drug_features.csv size:
n*dim(1900) - protein_features.csv size:
n*dim(1900)
- ddi_edgelist.csv :
idx idx - ddi_edgeweight.csv :
inter-scorce(float)
- ppi_edgelist.csv :
idx idx - ppi_edgeweight.csv :
inter-scorce(float)
- neg_test_idx.csv :
idx idx - neg_train_idx.csv :
idx idx - pos_test_idx.csv :
idx idx - pos_train_idx.csv :
idx idx
to-do list as below
- get protein and drug categories
- gen they embedding
- gen DTI DDI PPI pairs
1.Read org MTT
2.Modify MTT to multitask PPI DTI PPI
- how to design new loss_fn :(
$loss = l_{dti} + w_{ppi} * l_{ppi} + w_{ddi} * l_{ddi}$ - which task is main task ?
- data for PPI DDI need spite into test\train\val?
3.add adversaial task