A deep learning framework for gene ontology annotations with sequence- and network-based information

Fuhao Zhang, Hong Song, Min Zeng, Fang-Xiang Wu, Yaohang Li, Yi Pan, and Min Li*

School of Computer Science and Engineering, Central South University, China

We propose a deep learning framework called DeepGOA to predict protein functions with protein sequences and protein-protein interaction (PPI) networks.

Source Code: DeepGOA

Usage

You first need to download the source code and dataset . Then you should extract the dataset to the DeepGOA directory. If you want to run our saved model, you should download saved_model and save to the DeepGOA directory. You can run the DeepGOA.py file to train DeepGOA and the model will be saved in checkpoints path. Before you run the Predict_DeepGOA.py file, you should move the model to the saved_model path or change the load path. If you want to tune some hyper-parameters, you can change some values of hyper-parameters in config.py in utils folder.

Supplementary

We have provided the prediction of DeepGOA on the test dataset that is given as results of DeepGOA on benchmark test dataset.rar . And the prediction of human proteins is given as result of DeepGOA for Human specie.rar . The format of these files as follows:

Table The format of these files.

Protein entry	GO term	Prediction score	Class type
Q92543	GO:0044464	0.887	cc
Q8NBJ7	GO:0008152	0.939	bp