perCpG

perCpG is a shorthand used in epigenomics to describe analyses carried out at the level of individual CpG dinucleotides rather than on larger genomic regions. CpG sites, where a cytosine is followed by a guanine in the 5'→3' direction, are common targets of DNA methylation; measuring methylation at single-CpG resolution is possible with technologies such as whole-genome bisulfite sequencing and array-based platforms (e.g., Illumina Infinium).

A per-CpG approach reports methylation metrics for each site, commonly as beta values (fraction methylated between 0 and 1) or M-values (logit-transformed). Statistical testing can use linear models, mixed models, or count-based approaches (e.g., beta-binomial models for sequencing reads). Because many thousands to millions of sites are tested, correction for multiple comparisons (false discovery rate) is essential.

Advantages of per-CpG analyses include high resolution for detecting site-specific changes and the ability to annotate findings to exact genomic coordinates or regulatory elements. Limitations include increased noise and variability at single sites, uneven coverage in sequencing data, cell-type heterogeneity confounding, and challenges in biological interpretation compared with differentially methylated regions (DMRs). Per-CpG results are often integrated with region-based analyses, functional annotation, and downstream pathways to improve robustness and interpretability in studies such as epigenome-wide association studies, biomarker discovery, and cancer epigenetics.