MGAE-DC

This is the implementation code of MGAE-DC, a deep learning framework for predicting the synergistic effects of drug combinations through multi-channel graph autoencoders.

Requirements

Python 3.8 or higher
pandas 1.3.5 numpy 1.21.2 tensorflow 2.4.1

Datasets

1. O'Neil dataset
2. ALMANAC dataset
3. CLOUD datset
4. FORCINA datset

Training

###The embedding module python codes/get_oneil_mgaedc_representation.py -learning_rate 0.001 -epochs 10000 -embedding_dim 320 -drop_out 0.2 -weight_decay 0 -val_test_size 0.1
This script is used to extract the cell line-specific and common drug embeddings through multi-channel graph autoencoders in the embedding module.

Argument	Default	Description
learning_rate	0.001	Initial learning rate.
epochs	10000	The number of training epochs.
embedding_dim	320	The number of dimension for drug embeddings.
dropout	0.2	Dropout rate (1 - keep probability)
weight_decay	0	Weight for L2 loss on embedding matrix.
val_test_size	0.1	the rate of validation and test samples.

###The predictor module python codes/get_oneil_mgaedc.py -learning_rate 0.01 -epochs 500 -batch 320 -drop_out 0.2 -hidden 8192 -patience 100 This script is used to predict the synergistic effects of drug combinations in the predictore module.

Argument	Default	Description
learning_rate	0.01	Initial learning rate.
epochs	500	The number of training epochs.
batch	256	The nbatch size.
hidden	0.2	Dropout rate (1 - keep probability)
weight_decay	1024	(n, n/2, 1) The hidden size for NN.
patience	100	the patience for early stop.

###Predicting with pretrained model

The size of pretrained models are too large, so they are accessible with baidu netdisk links. 密码：bkri.

Reference

Please cite our work if you find our code/paper is useful to your work.

@article{Zhang, 
title={MGAE-DC: predicting the synergistic effects of drug combinations through multi-channel graph autoencoders}, 
author={Peng Zhang, Shikui Tu}, 
journal={}, 
volume={}, 
number={}, 
year={2022}, 
month={}, 
pages={} 
}