MOGONET : Multi-omics Integration via Graph Convolutional Networks for Biomedical Data Classification
MOGONET integrates multi-omics data using graph convolutional networks
MOGONET combines GCN for multi-omics-specific learning and VCDN for multi-omics integration. For clear and concise illustration, an example of one sample is chosen to demonstrate the VCDN component for multi-omics integration. Preprocessing is first performed on each omics data type to remove noise and redundant features. Each omics-specific GCN is trained to perform class prediction using omics features and the corresponding sample similarity network generated from the omics data. The cross-omics discovery tensor is calculated from the initial predictions of omics-specific GCNs and forwarded to VCDN for final prediction. MOGONET is an end-to-end model and all networks are trained jointly. Here is the original article et github.
However, the model can be trained using only two types of omics instead of 3 or more, depending on your data.
main_mogonet.py: Examples of MOGONET for classification tasks
MOGONET.py: How to create your data et save in to folder
main_biomarker.py: Examples for identifying biomarkers
models.py: MOGONET model
train_test.py: Training and testing functions
feat_importance.py: Feature importance functions
utils.py: Supporting functions