A Smoothing Method For DNA Methylome Analysis To Enhance Epigenomic Signature Detection In Epigenome-Wide Association Studies
Epigenome-wide association studies (EWAS) are vital for uncovering how changes in DNA methylation relate to human traits and diseases. However, these studies frequently face hurdles like low statistical power and a tendency to produce false positive results, particularly in smaller research cohorts.
To address these limitations, a novel computational method has been developed. This technique enhances the analysis of DNA methylation data by leveraging the natural tendency of adjacent DNA methylation sites, known as CpG dinucleotides within CpG islands, to exhibit similar patterns. It employs a “smoothing” approach, which essentially involves calculating a sliding-window average of methylation levels, and further refines this process using Savitzky-Golay filtering.
The benefits of this innovative method are substantial. It significantly improves the signal-to-noise ratio, making true biological signals stand out more clearly, while simultaneously reducing background noise and the likelihood of spurious findings. This allows for the more accurate identification of important epigenetic changes, even in studies with limited sample sizes. For instance, in one application, it boosted the signal-to-noise ratio by 90% and cut noise variance by 80% at a specific target region. The method also proved effective in deriving an “Epigenetic CpG aging score” that showed a strong correlation with chronological age. This advancement offers a valuable tool for re-analyzing existing methylation datasets to uncover previously undetected epigenomic insights.
Source: link to paper