npGSEA

npGSEA performs parametric gene set enrichment analysis by approximating permutation distributions for the sums and sums of squared correlations test statistics to enable efficient evaluation of gene-set associations in high-throughput expression data.

Key Features:

• Parametric approximation: Uses a moment-based approach to approximate permutation distributions of test statistics, reducing the number of permutations required to obtain small p-values.
• Supported test statistics: Provides parametric approximations specifically for the sums and sums of squared correlations gene set test statistics.
• Efficiency and computational cost: For linear gene set tests the effective computational cost is approximately equivalent to |G| permutations and for quadratic statistics on the order of |G|^2 permutations, representing large reductions relative to exhaustive permutation sampling.
• Accuracy and ranking preservation: Parametric approximations produce p-values and gene-set rankings that closely match exhaustive permutation results.
• Affiliation: Developed as part of the Bioconductor project.
• Application to empirical datasets: Demonstrated on three public Parkinson’s Disease expression datasets where it identified enriched gene sets not previously reported.

Scientific Applications:

• High-throughput expression analysis: Testing relationships between gene collections and outcomes in microarray or RNA-seq expression studies.
• Multiple simultaneous tests: Efficiently handling large numbers of gene-set tests common in genomics and systems biology.
• Neurodegenerative disease studies: Identification of enriched gene sets in Parkinson’s Disease expression datasets.

Methodology:

Calculate exact moments of the test statistics (sums and sums of squared correlations) under the permutation null, fit moment-based parametric distributions to those moments to approximate permutation distributions, and validate approximations via simulation comparisons to permutation testing (Ackermann and Strimmer, 2009).