MOSS

High dimensionality, noise and heterogeneity among samples and features challenge the omic integration task. Here we present an omic integration method based on sparse singular value decomposition (SVD) to deal with these limitations, by: a. obtaining the main axes of variation of the combined omics, b. imposing sparsity constraints at both subjects (rows) and features (columns) levels using Elastic Net type of shrinkage, and c. allowing both linear and non-linear projections (via t-Stochastic Neighbor Embedding) of the omic data to detect clusters in very convoluted data (Gonzalez-Reymundez et. al, 2022) doi:10.1093/bioinformatics/btac179.