CIADTI

we propose an end to end deep learning framework to extract robust intermolecular and intramolecular features of drugs and targets, and predict potential DTIs, named CIADTI. Three parallel modules are proposed, that one module is utilized to extract intermolecular features between drugs and targets, while another two modules are used to produce intramolecular features for drugs and targets, respectively. Finally, all features are concatenated and fed into fully connected dense layers for predicting DTIs.

Requirements

• python >= 3.5
• torch >= 1.4.0
• RDkit >= 2019.03.30
• gensim >= 3.4.0
• numpy >= 1.16.1
• pandas >= 1.1.4
• transformers >= 3.1.0

Using

1. model.py: the construction of the neural network
2. data_preprocess: Process the data to get the input of the model
3. main.py: start file for model training