Analysis of genotype by methylation interactions through sparsity-inducing regularized regression

Area of Science:

• Genetics
• Epigenetics
• Statistical Genomics

Background:

• Genome-wide association studies (GWAS) and epigenome-wide association studies (EWAS) are crucial for identifying genetic and epigenetic factors influencing complex traits.
• LASSO-type regression techniques offer powerful tools for variable selection in high-dimensional genomic data.

Purpose of the Study:

• To adapt LASSO-type regression for detecting important genetic and epigenetic loci in GWAS and EWAS.
• To incorporate quantifiable uncertainty and control family-wise error rates using stability selection.
• To explore genetic and epigenetic interactions using group LASSO variants.

Main Methods:

• Application of LASSO-type regression techniques, including group LASSO.
• Utilizing stability selection for quantifiable uncertainty and family-wise error rate control.
• Analysis of the Genetics of Lipid Lowering Drugs and Diet Network (GOLDN) dataset, including CpG sites and SNPs.

Main Results:

• Successful replication of existing findings for 2 CpGs in carnitine palmitoyltransferase 1A (CPT1A) using the GOLDN dataset.
• Identification of potential new research directions for analyzing genetic and epigenetic interactions.
• Demonstration of LASSO-based genome-wide and epigenome-wide models, including a chromosome 11 interaction model.

Conclusions:

• LASSO-type regression, enhanced with stability selection, provides a robust framework for genetic and epigenetic association studies.
• The methods facilitate the detection of significant loci and interactions with controlled error rates.
• Further exploration of genetic and epigenetic interactions is warranted.