This work is under new modifications, it will be updated as soon as the results will be reliable. Thank you>>> This repository is part of a work for classifying interaction between compound represented by their chemical sequence and proteins also a sequential data. The aim of this is to integrate Graph for representing the chemical structure and CNN combined with recurrent layers for to obtain feature vector of the protein. The result is a binary class of the interaction or non interaction between the molecule and the protein.
Data is obtained from the script as given in the "Notice.txt" and need to emphasize again that data is desigend from Liu et al "Improving compound–protein interaction prediction by building up highly credible negative samples (Bioinformatics, 2015)." and that the ratio of positive and negative samples is 1:1.
The architectura of the model the reprssent this code integrate of several deep learnng models model such as Graphs, Convolution Neural Networks followed Bi-Recurrent layers on to learn chemical and protein sequences and improve the accuracy of the prediction of their interections.
#Main Packgages
- PyTorch
- scikit-learn
- RDKit
This repository will be upated again with more material upon the paper, which this code is used for, will be published to a repsective journal. Also and extended result review for implementing and comparing with more machine learning methods will be provided in a future moment.
This work is based on previoues model "Notice.txt" used and modified some script and with the intention for more updated in future for further more explations after the paper will be published in the selected Journal.