CoGAPS

CoGAPS infers coordinated gene-set activities from transcriptomic data by isolating gene expression patterns driven by biological processes.

Key Features:

• Bayesian MCMC matrix factorization (GAPS): Uses the GAPS Bayesian Markov chain Monte Carlo matrix factorization algorithm to perform sparse decomposition of high-dimensional transcriptomic data into interpretable patterns.
• Threshold-independent gene set analysis: Applies a threshold-independent statistic to infer activity of predefined gene sets without relying on arbitrary cutoffs.
• Sparse pattern representation: Produces sparse factor matrices that represent coordinated gene activity associated with specific biological processes.
• Implementation on JAGS: Implements the GAPS algorithm as C++ code built on top of JAGS (Just Another Gibbs Sampler) for Bayesian inference.

Scientific Applications:

• Cancer biology: Deconvolves coordinated gene activity in cancer transcriptomes to elucidate underlying biological mechanisms.
• Genomics: Detects process-level signals in high-throughput transcriptomic datasets for genomics analyses.
• Systems biology: Characterizes coordinated pathway or gene-set activities for systems-level modeling.
• Disease modeling and biomarker discovery: Reveals gene-set activity patterns that support disease modeling and biomarker identification from transcriptomic data.
• Therapeutic target identification: Identifies process-associated gene expression patterns that can inform candidate therapeutic targets.

Methodology:

Performs Bayesian MCMC matrix factorization using the GAPS algorithm implemented in C++ on top of JAGS and computes a threshold-independent gene set activity statistic on the inferred patterns.