Poly-Enrich: count-based methods for gene set enrichment testing with genomic regions

Area of Science:

• Genomics and Bioinformatics
• Molecular Biology

Background:

• Gene set enrichment (GSE) analysis is crucial for interpreting large genomic datasets like ChIP-seq.
• Existing methods (ChIP-Enrich, Broad-Enrich) have limitations in applicability and modeling complex genomic region data.

Purpose of the Study:

• To introduce Poly-Enrich, a novel GSE method with broader applicability and enhanced capabilities for genomic regions.
• To improve the biological interpretation of ChIP-seq and similar genomic region datasets.

Main Methods:

• Poly-Enrich utilizes a generalized additive model with a negative binomial family to model the number of peaks assigned to a gene.
• The method incorporates adjustments for gene locus length and handles scenarios where nearly all genes have associated peaks.
• Comparison with ChIP-Enrich using ENCODE ChIP-seq data to evaluate performance across different pathways and transcription factors.

Main Results:

• Poly-Enrich demonstrates wider applicability and improved performance, particularly for characterizing pathways and genic regions enriched with repetitive elements.
• The optimal GSE test is influenced more by the regulated pathway than the transcription factor's properties.
• Certain biological processes, potentially involving multiple binding events, are better modeled by Poly-Enrich's count-based approach.

Conclusions:

• Poly-Enrich offers a more robust and versatile method for gene set enrichment analysis of genomic regions.
• The findings suggest that complex regulatory mechanisms involving multiple binding events are better captured by count-based GSE methods.
• A new hybrid method automatically selects the optimal testing approach for each gene set with appropriate FDR adjustment.