Penalized-regression-based multimarker genotype analysis of Genetic Analysis Workshop 17 data

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Association studies typically analyze single markers, potentially missing effects of multiple variants.
• Untyped causal variants in linkage disequilibrium with typed markers can be captured by multimarker analysis.
• Identifying the combined effect of multiple markers is crucial for comprehensive genetic analysis.

Purpose of the Study:

• To propose and evaluate a novel sliding window approach using multimarker genotypes in penalized regression.
• To investigate a penalty with three components: group LASSO, allele-sharing, and coefficient shrinkage.
• To compare the proposed method with single-marker analysis and gene-based sparse group LASSO.

Main Methods:

• A sliding window approach utilizing multimarker genotypes as variables.
• A penalized regression model incorporating group LASSO, allele-sharing, and shrinkage penalties.
• Minimization of penalized likelihood using a cyclic coordinate descent algorithm.
• Comparison with single-marker analysis and sparse group LASSO on Genetic Analysis Workshop 17 data for quantitative trait Q2.

Main Results:

• All tested methods, including the proposed one, were underpowered to detect simulated rare causal variants at desired low false-positive rates.
• The sparse group LASSO applied to multi-marker genotypes showed a potential advantage over its application to single nucleotide polymorphisms (SNPs) within genes.
• Evidence suggests that modeling combinations of rare variant alleles may be more advantageous than individual modeling.

Conclusions:

• The proposed penalized regression method offers a way to model multimarker genotypes.
• While current methods struggle with rare variants, modeling allele combinations shows promise.
• Further research is needed to enhance power for detecting rare variants in association studies.