GNE: A deep learning framework for gene network inference by aggregating biological information

This is the tensorflow implementation of the GNE as described in our paper.

GNE integrates gene interaction network with gene expression data to learn a more informative representations for gene network, which can be plugged into off-the-shelf machine learning methods for diverse functional inference tasks: gene function prediction, gene ontology reconstruction, and genetic interaction prediction.

Architecture of GNE

Requirements

Example to run demo.

python run_GNE.py

Data

In order to use your own data, you have to provide

We load interaction network data and expression data matrix for yeast in run_GNE.py as an example. The original datasets can be found from data sources:

Dataset Source
Interaction dataset BioGRID
Gene expression data DREAM5 Challenge

You can use M by M adjacency matrix as input by editing run_GNE.py.

Note: the order of genes in adjacency matrix and feature matrix should be same.

Settings

You can choose between the following settings:

For coding, We name λ as alpha in parameters list.

Supplementary Materials

Supplementary Table S1
Supplementary Table S2

For more details, visit GNE references.

Contact

kk3671@rit.edu