What is it about?
DNA methylation (DNAm) is a major epigenetic mechanism that can change the activity of a DNA segment without changing its sequence by adding a methyl (CH3) group to the cytosine base. DNAm plays important roles in gene regulation, brain development, and function. Genetic variations may impact DNAm levels in the human brain and further contribute to the genetic basis of brain disorders. Population-based genetic association studies have identified numerous DNAm quantitative trait loci (mQTLs) associated with DNAm levels at particular CpG sites, but due to extensive linkage disequilibrium (LD) across the genome, it is challenging to identify specific genetic variations that drive DNAm levels of CpG sites, which may limit the utility of mQTLs for pinpointing casual variants within GWAS risk loci. In this study, we designed a deep learning model INTERACT that is able to identify genetic variations that drive DNAm levels in the human brain, but do not suffer from LD confounding effect.
Featured Image
Photo by Sangharsh Lohakare on Unsplash
Why is it important?
Our INTERACT model integrates the classical convolutional neuronal network with the state-of-the-art transformer model and has the strength of detecting both local and distant interacting features, such as DNA motifs. Additionally, the pretraining and fine-tuning design enables knowledge transfer from whole-genome bisulfite sequencing dataset to EPIC array dataset, leveraging strengths and limitations of both datasets.
Perspectives
Read the Original
This page is a summary of: Deep learning predicts DNA methylation regulatory variants in the human brain and elucidates the genetics of psychiatric disorders, Proceedings of the National Academy of Sciences, August 2022, Proceedings of the National Academy of Sciences,
DOI: 10.1073/pnas.2206069119.
You can read the full text:
Contributors
The following have contributed to this page