bootGSEA: a bootstrap and rank aggregation pipeline for multi-study and multi-omics enrichment analyses

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Gene set enrichment analysis (GSEA) is standard in omics data analysis but faces reproducibility challenges due to dynamic database annotations.
• Changes in gene or protein set compositions can significantly impact biological interpretations derived from GSEA.

Purpose of the Study:

• To introduce bootGSEA, a novel computational pipeline for assessing the robustness and variability of GSEA results.
• To enhance the reliability and reproducibility of GSEA by accounting for changes in set annotations.

Main Methods:

• bootGSEA employs bootstrap sampling to generate multiple GSEA replicates, varying gene/protein inclusion in each.
• Rank aggregation is utilized to combine results from bootstrap replicates and compare them to standard GSEA.
• The pipeline facilitates combining results across different omics levels or multiple studies.

Main Results:

• Application to cancer transcriptomics datasets demonstrated that bootstrap GSEA identifies more robustly enriched gene sets.
• Analysis of paired transcriptomics and proteomics data from a spinal muscular atrophy mouse model showed robust rankings at both omics levels.
• The bootGSEA R-package was developed, offering graphical visualizations and identifying less robust gene/protein sets under varying compositions.

Conclusions:

• Bootstrap-based GSEA enhances the selection of reliable gene sets, mitigating issues from changing database annotations.
• Rank aggregation effectively integrates bootstrap results for robust single-omics or multi-omics findings.
• The bootGSEA pipeline and R-package provide a valuable tool for reproducible and robust GSEA in omics research.