NETWORK-BASED MULTI-OMICS INTEGRATION TO PRIORITIZE FEATURES IN GEMMA OMICS DATASETS

NETWORK-BASED MULTI-OMICS INTEGRATION TO PRIORITIZE FEATURES IN GEMMA OMICS DATASETS

We developed a pipeline for the integrative analysis of gene-related, metabolite-related and microbiota species-related data. The pipeline relies on a network of molecular interactions and functional interactions (interactome), and on the network analysis tool “mND” (multi-layer Network Diffusion) [doi: 10.1093/bioinformatics/btz652].

Gene-related, metabolite-related and microbiota species-related input scores from GEMMA omics are mapped on the interactome (see “A network of molecular and functional interactions for GEMMA multi-omics data analysis”), and mND is used to prioritize features (genes, metabolites and microbial species), based on their input scores and reciprocal network proximity. The output feature-level score is statistically assessed based on two distinct null models, which test the relation between output scores and input scores, and between output scores and feature-feature interactions.

The top scoring features are selected based on feature-level score, network topology scores (e.g. modularity, connected components), and joint analysis of the two (e.g. distribution of feature score over the network).