The Smoothing Method for DNA Methylome Analysis Identifies Highly Accurate Epigenomic Signatures in Epigenome-Wide Association Studies

Abstract

Abstract Background: The genome-wide assessment of the DNA methylome has revolutionized our comprehension of epigenome alterations linked to complex human traits and diseases. The ability of epigenome-wide association studies (EWAS) to translate into biologically meaningful results relies on detecting epigenomic signatures with a high level of statistical certainty. However, the classical analyses of EWAS are prone to statistical inflation and bias, leading to spurious associations, particularly in case series with small sample sizes, such as those analyzing patients with rare inherited disorders. Methods: Based on the co-methylation pattern of CpG dinucleotides within the CpG islands, we propose the smoothing method at the genome-wide level through a sliding window approach to calculate and visualize data from EWAS to decipher the most informative epigenetic alterations of EWAS with a high degree of accuracy. Results: The smoothing method is a simple method that identifies epigenomic signatures with a high degree of certainty while controlling the risk of spurious findings outside the significant loci at a genome-wide level. We have systematically compared the smoothing method with a classical supervised approach in several EWAS settings, including two monogenic epigenetic diseases (epi-cblC and primary constitutional MLH1epimutation) and epigenetic predictors of aging. In the latter example, we showed that the smoothing method remained efficient even after applying an 80% reduction of the original sample size. Conclusions: The smoothing method for DNA methylation analyses is based on the biological correlate of the epigenome structure and identifies highly accurate epigenomic signatures in DNA methylation analyses. Its application to several settings of epigenome-wide analyses confirmed its usefulness for deciphering the most informative epigenomic signatures with a high degree of certainty while controlling the risk of spurious findings outside the significant loci at a genome-wide level. Our results suggest revisiting EWAS by applying the smoothing method to already available datasets to re-analyze and potentially identify highly accurate epigenomic signatures that could translate into biologically meaningful results.