Differential network-based methods for the integration of omics data: overview and challenges

The focus of biomedical research has shifted from single-omics data analysis to the concurrent use of multi-omics datasets in order to obtain accurate results and a better understanding of complex diseases, which frequently display perturbations in multiple omics layers. Although omics data can be examined individually and the resulting findings matched, breakthrough methods that allow for the systematic integration of omic layers are promptly needed.

Recently, many machine learning-based techniques have approached high-order integration of multi-omics data, mostly for predictive purposes. Their findings, however, can only provide a weak mechanistic explanation for what is happening at the system level. Within the framework of statistical models, network-based multi-omics integration retrieves relationships between biomolecules in and across different omic layers. Differential networking techniques detect differences in the pairwise interactions of biomolecules between conditions. The use of network-based multi-omics integration methods in conjunction with differential networking may aid in the discovery of disease-specific multi-omics features.

Differential multi-omics networks provide a better understanding of the interactome rewiring that underpins disease start and progression. The examination of disease modules enables the discovery of novel molecular traits and genotype-phenotype associations. Furthermore, they enable the use of network-based methodologies for the identification of prospective biomarkers and therapy candidates as well as possible combination therapy approaches to provide better efficacy and lower toxicity.