Deconvolution of various types of omics data using matrix factorization approaches

Characterization of a tumor nowadays implies various omics experiments (transcriptomics, methylation, proteomics, genomics, ...). Tumor complexity can be assessed either in an integrative way combining all information or layer by layer. Here we performed and compared both, using matrix factorization, on one the hand a multi-omics deconvolution method: Multi-Omics Factor Analysis (MOFA) and on the other hand, Independent Component Analysis (ICA) applied to each layer. Ongoing preliminary results show that both approaches give insightful results and show similarities as well as specificities.