wanglx1874 / gonet Goto Github PK

This repository contains implementation of GONET in jupyter notebook

You should download datasets mentioned in our paper, and use the scripts to train the model.

Model details:

sequence data: mapping to matrix using ProVec ,i.e. each trigram is represented by a 100d dense vector, so each protein sequnce is mapped to a 1500*100 matrix. feature extraction of sequence data: recurrent CNNS followed by attention.

PPI data: utilize network representation learning to get embeddings of each protein node, concat this vector with sequence feature to predict labels.

Important: i write the codes in different machines, so the datapathes in the scripts may be incorrect,please update the path with respect to the true path. what is more, hyper-parameters in our model is not the best, which means you can change them to get better results.