Gene-set analysis (GSA) summarizes information from molecule to gene-set level to facilitate biological interpretation of molecular profiling experiments. We present a statistical framework for single sample GSA of multiple 'omic data. MOGSA learns the most variant biomolecules and integrates data sets to generate gene set scores (GSS) for each sample. It extracts the contribution of individual data sets and biomolecules to GSS. MOGSA is rigorously benchmarked with simulated and real biological data and is implemented in the Bioconductor package “mogsa.”