mGSZ

mGSZ applies the Gene Set Z-score (GSZ) scoring function with asymptotic P-value estimation to perform gene set analysis that evaluates the collective effects of gene sets on biological processes.

Key Features:

• Asymptotic P-Value Calculation: Uses asymptotic P-values instead of extensive permutation-based empirical P-values to reduce the need for numerous permutations and accelerate computation.
• GSZ Scoring Function: Centralizes the GSZ scoring function, which was demonstrated to outperform seven other popular gene set scoring functions in comparative studies.
• Max-Mean Statistics Replacement: Replaces max-mean statistics with the GSZ scoring function to enhance gene set analysis performance.
• Permutation Importance: Emphasizes the necessity of both gene and sample permutations for accurate P-value estimation and highlights risks when either permutation is omitted.
• Comparison with Empirical Methods: Employs an asymptotic approach that offers advantages over empirical methods in efficiency and precision of P-value estimation.
• Performance Evaluation: Demonstrates superior performance in two distinct evaluations and reports that rotation tests do not improve the asymptotic P-values from its methodology.
• Asymptotic Distribution Models: Proposes well-known asymptotic distribution models for three of the compared gene set analysis methods.
• R Package Implementation: Provides an implementation as an R package.

Scientific Applications:

• Genomics: Enables gene set analysis for genomic studies to assess coordinated gene effects on biological processes.
• Transcriptomics: Supports analysis of transcriptomic data to detect pathway-level changes using GSZ scoring and asymptotic P-values.
• Systems Biology: Facilitates systems-level interpretation of gene set effects within biological networks and processes.
• High-Throughput Studies: Suited for high-throughput datasets where reduced computational demands for P-value estimation are beneficial.

Methodology:

Applies the GSZ scoring function and computes asymptotic P-values rather than empirical permutation P-values; substitutes max-mean statistics with GSZ; highlights the use of both gene and sample permutations for P-value accuracy; compares GSZ against seven other scoring functions and evaluates performance in two distinct tests, reporting that rotation tests do not improve asymptotic P-values; proposes asymptotic distribution models for three compared methods; implemented as an R package.