rapidGSEA

RapidGSEA: Accelerated Gene Set Enrichment Analysis via GPU and Multi-core Parallelization

RapidGSEA accelerates Gene Set Enrichment Analysis (GSEA) by computing enrichment scores and estimating their statistical significance using parallelized permutation testing on GPUs and multi-core CPUs. It evaluates associations between predefined gene sets and phenotypes by calculating a weighted running sum statistic across ranked gene expression profiles.

Key Features:

• cudaGSEA: Implements fine-grained parallelization using CUDA for GPU acceleration; processes 4,725 gene sets from 20,639 unique gene symbols across 200 patients (183 cases, 17 controls) in under one hour, achieving nearly two orders-of-magnitude speedup over broadGSEA.
• ompGSEA: Uses coarse-grained multi-threaded parallelization on multi-core Xeon CPUs; achieves approximately one order-of-magnitude speedup over broadGSEA in CPU-only environments.
• Permutation-based Significance Testing: Performs non-parametric permutation of phenotype labels to estimate enrichment score null distributions and compute p-values under multiple hypothesis testing.

Scientific Applications:

• Phenotype-Associated Gene Set Identification: Detects significantly enriched predefined gene sets based on differential gene expression between sample groups.

Methodology:

Ranks genes by differential expression between phenotypic groups, computes enrichment scores as weighted running sum statistics across ranked genes, and estimates statistical significance through repeated non-parametric permutation of phenotype labels to approximate null distributions for multiple gene sets.