A machine learning approach to brain epigenetic analysis reveals kinases associated with Alzheimer's disease

Alzheimer's disease (AD) is influenced by both genetic and environmental factors; thus, brain epigenomic alterations may provide insights into AD pathogenesis. Multiple array-based Epigenome-Wide Association Studies (EWASs) have identified robust brain methylation changes in AD; however, array-based assays only test about 2% of all CpG sites in the genome. Here, we develop EWASplus, a computational method that uses a supervised machine learning strategy to extend EWAS coverage to the entire genome. Application to six AD-related traits predicts hundreds of new significant brain CpGs associated with AD, some of which are further validated experimentally. EWASplus also performs well on data collected from independent cohorts and different brain regions. Genes found near top EWASplus loci are enriched for kinases and for genes with evidence for physical interactions with known AD genes. In this work, we show that EWASplus implicates additional epigenetic loci for AD that are not found using array-based AD EWASs. Array-based epigenome-wide association studies only test about 2% of the CpG sites in the genome. Here, the authors describe EWASplus, a supervised machine learning strategy that extends EWAS coverage to the entire genome, and use it to identify novel brain CpGs associated with Alzheimer's disease.

Automated summary

The study introduces a method called EWASplus that extends the coverage of epigenome-wide association studies (EWAS) to the entire genome using supervised machine learning, and identifies novel brain CpG sites associated with Alzheimer's disease.