MultiNEP: disease-specific Multi-omics Network Enhancement for Prioritizing disease genes and metabo

Zhuoran Xu, Brian Lee, Luigi Marchionni, Shuang Wang. Weill Cornell Medicine

Many studies have used different types of omics profiles together with appropriate networks to prioritize disease-associated genes, DNA methylation sites, or metabolites. Metabolome, as the last layer before phenotypes and the end-product of gene expressions and protein activities, has accumulated growing attentions. Combining metabolome and transcriptome to simultaneously prioritize disease-associated metabolite abundances and gene expressions using a "multi-omics network" being constructed with three individual networks (a gene-gene network, a metabolite-metabolite network, and a gene-metabolite network) could further utilize interaction information between metabolites and gene expressions that are not used when prioritizing them separately. However, the number of metabolites is usually 100 times smaller than that of genes, without accounting for this imbalance issue, additional information in the gene-metabolite network might be buried in the amount of information in a gene-gene network when prioritizing gene expressions, while gene-metabolite interactions might completely dominate metabolite-metabolite interactions when prioritizing metabolites. No methods exist that focus on simultaneously prioritizing gene expressions and metabolites using networks. Here we developed a disease-specific Multi-omics Network Enhancement Prioritization (MultiNEP) framework that employs two weighting parameters to control contributions of individual networks within a "multi-omics network" that handles imbalance to simultaneously prioritize disease-associated gene expressions and metabolites. Built upon our previous work on a single network, we first enhanced a general multi-omics network into a disease-specific multi-omics network using multiple types of disease omics. In both simulation studies and real data applications, when we down-weight gene-gene network's and up-weight metabolite-metabolite network's relative contribution to gene-metabolite network, MultiNEP outperforms competing methods including methods not address network imbalance and methods only use one type of omics data.