Meta-analysis approaches to combine multiple gene set enrichment studies

Area of Science:

• Bioinformatics
• Computational Biology
• Statistical Genetics

Background:

• Meta-analysis integrates multiple studies for reliable biomedical research summarization and increased power in detecting gene sets for complex diseases.
• Existing methods like Meta-Analysis for Pathway Enrichment (MAPE) may suffer power loss due to summary statistics and set-size dependent enrichment scores.

Purpose of the Study:

• To adapt genome-wide association study meta-analysis approaches (fixed and random effects models) for integrating multiple gene set enrichment analysis (GSEA) studies.
• To develop a mixed strategy with adaptive testing for selecting between random effects (RE) and fixed effect (FE) models for improved statistical efficiency and flexibility.
• To propose a size-adjusted enrichment score using a one-sided Kolmogorov-Smirnov statistic to account for varying gene set sizes.

Main Methods:

• Adaptation of fixed effect and random effects meta-analysis models from genome-wide association studies for GSEA data integration.
• Development of a mixed strategy employing adaptive testing to dynamically choose between RE and FE models.
• Introduction of a size-adjusted enrichment score based on the one-sided Kolmogorov-Smirnov statistic.

Main Results:

• The proposed methods demonstrate significantly better performance compared to existing MAPE methods.
• The adaptive testing strategy shows the most stable performance across various general situations.
• The methods are applicable to both discrete and continuous phenotypes.

Conclusions:

• The developed meta-analysis approaches enhance the integration of GSEA studies, leading to more powerful and reliable detection of disease-associated gene sets.
• Adaptive testing and size-adjusted enrichment scores address limitations of previous methods, offering improved statistical efficiency and flexibility.
• These advanced methods provide a robust framework for analyzing complex human diseases through gene set enrichment analysis.