Gene-set analysis is severely biased when applied to genome-wide methylation data

Area of Science:

• Epigenetics
• Genomics
• Bioinformatics

Background:

• DNA methylation is a key epigenetic regulator of gene expression with significant biological and clinical implications.
• Genome-wide methylation assays are crucial for comparing methylation patterns across sample groups.
• Gene set analysis is commonly used to identify pathways affected by differential DNA methylation in diseases.

Purpose of the Study:

• To identify and address the bias in gene set analysis of genome-wide methylation data.
• To demonstrate the impact of this bias on identifying differentially methylated genes and pathways.
• To propose and validate methods for correcting this bias to enable more accurate biological insights.

Main Methods:

• Analysis of published lung cancer methylation data and a newly generated ulcerative colitis dataset.
• Application of gene set analysis to identify enriched gene sets in methylation data.
• Utilizing randomized data to assess the significance of identified gene sets and quantify bias.
• Adapting existing statistical approaches to correct for the identified bias.

Main Results:

• Gene set analysis of genome-wide methylation data is severely biased due to variations in CpG site numbers per gene.
• Enrichment results in previous studies may be artifacts of this bias, as demonstrated with randomized data.
• Correcting the bias in lung cancer and ulcerative colitis datasets revealed novel insights into methylation's role in disease.
• Many previous genome-wide methylation studies may have drawn inaccurate conclusions due to this analytical bias.

Conclusions:

• Standard gene set analysis methods applied to genome-wide methylation data are inherently biased.
• This bias can lead to spurious findings and misinterpretation of biological pathways.
• Correcting for this bias is essential for accurate interpretation of methylation data and advancing our understanding of diseases like cancer and chronic inflammation.