mogsa

mogsa integrates multiple omics datasets to compute integrated gene set scores and reveal pathway-level signals across transcriptomics, proteomics, copy number variation, and other molecular data.

Key Features:

• Integration Across Multiple Omics Data Types: Integrates diverse molecular data such as transcriptomics and proteomics without requiring the intersection of features (e.g., gene IDs), allowing inclusion of unannotated features.
• Low-Dimensional Representation Learning: Identifies a low-dimensional representation of the most variant correlated features across datasets and projects features onto a common scale to derive integrated gene set scores.
• Enhanced Analytical Power and Noise Reduction: Increases power to detect subtle gene-set changes and mitigates dataset-specific noise by leveraging multiple molecular layers.
• Assessment of Data Type Influence: Evaluates the contribution of each data type or individual features to the overall gene set score.

Scientific Applications:

• Noise reduction in omics data: Demonstrated reduction of noise using NCI60 transcriptome and proteome data.
• Comparative analysis of stem cell profiles: Compared mRNA, protein, and phosphorylation profiles between induced pluripotent stem cells (iPSCs) and embryonic stem cell lines to reveal similarities and differences.
• Cancer subtype discovery: Identified three robust molecular subtypes in bladder cancer samples from The Cancer Genome Atlas by integrating copy number variation and mRNA profiling.

Methodology:

Applies a multivariate approach that identifies low-dimensional representations of the most variant correlated features, transforms features onto a common scale, computes integrated gene set scores, and assesses per-data-type contributions.