The challenge of detecting genotype-by-methylation interaction: GAW20

Area of Science:

• Genetics
• Epigenetics
• Bioinformatics

Background:

• GAW20 working group 5 convened researchers to evaluate methods for detecting genetic by epigenetic interactions.
• Utilized real data from the Genetics of Lipid Lowering Drugs and Diet Network (GOLDN) study, including single-nucleotide polymorphism (SNP) markers, methylation (cytosine-phosphate-guanine [CpG]) markers, and phenotype data.
• A simulated dataset based on the GOLDN study data was also provided for analysis.

Purpose of the Study:

• To assess the performance of various statistical methods in identifying genetic by epigenetic interactions.
• To analyze both real-world genetic and epigenetic data and simulated data for robust evaluation.

Main Methods:

• Employed diverse statistical approaches including generalized linear mixed models, mediation analysis, machine learning, W-test, and sparsity-inducing regularized regression.
• Analyzed single-nucleotide polymorphism (SNP) markers, cytosine-phosphate-guanine (CpG) methylation sites, and their interactions.
• Utilized both real data from the GOLDN study and a simulated dataset.

Main Results:

• Statistical methods generally performed well in detecting genetic by epigenetic interactions.
• Confirmed causative SNPs on chromosome 11 in both simulation and real data.
• Identified significant findings for various SNPs, CpG sites, and SNP-CpG interaction pairs.

Conclusions:

• The study demonstrated good statistical power for detecting large genetic by epigenetic interaction effects.
• Detecting smaller interaction effects necessitates larger sample sizes or collaborative research efforts (consortiums).
• Further research is needed to enhance power for identifying subtle genetic and epigenetic associations.