Array probe density and pathobiological relevant CpG calling bias in human disease and physiological DNA methylation...

Area of Science:

• Epigenomics
• Genomics
• Bioinformatics

Background:

• The HumanMethylation450 BeadChip array (450K) is a key tool in epigenomics research.
• A known issue with the 450K platform is how the number of probes per gene (PG) affects the ranking of differentially methylated CpGs (DM-CpGs).
• Previous work indicated that ranking DM-CpGs by the ratio of DM-CpGs to PG enriched pathways overlapping with expression data.

Purpose of the Study:

• To evaluate the impact of probe count per gene (PG) on differentially methylated CpG (DM-CpG) identification using the HumanMethylation450 BeadChip array.
• To compare a novel ranking method (DM-CpGs-to-PG ratio) against standard methods (delta-beta ranking, gometh pipeline).
• To assess the biological relevance and discoverability of DM-CpGs identified through different ranking strategies.

Main Methods:

• Comparative analysis of thirteen 450K-based epigenomic studies across various biological contexts (FA-stimulated models, aging, disease, normal tissues).
• Ranking differentially methylated CpGs (DM-CpGs) using the ratio of DM-CpGs to the number of probes per gene (PG).
• Evaluation of gene ontology category enrichment and overlap with gene expression data for top-ranking DM-CpGs.

Main Results:

• The 150 top-ranking DM-CpGs are predominantly located in genes with a high number of probes per gene (PG).
• Delta-beta-based ranking shows a significant false-negative rate in genes with low PG.
• Ranking by DM-CpGs-to-PG ratio yields significantly enriched, biologically relevant gene categories and improves overlap with expression data compared to other methods.
• The top 15 DM-CpG loci are enriched in non-coding RNAs, a transcript type often underrepresented in 450K data.

Conclusions:

• A simple normalization method using the DM-CpGs-to-PG ratio effectively identifies pathobiologically relevant DM-CpGs.
• This method enhances the discovery of biologically meaningful signals from 450K array data.
• The proposed method is also applicable to the newer MethylationEPIC array, improving its utility in epigenomic research.