Methods of association for genome data with rare variants and a multinomial response

Abstract

Scope and Method of Study:

A rare variant is a Single Nucleotide Polymorphism (SNP) with a minor allele frequency (MAF) of 5% or less. Approximately 60% of human SNPs are rare variants. New rapid genotyping technologies now make it possible to efficiently survey these rare variants. Many new statistical methods are being developed to analyze the associations between rare variants and phenotypes. Current methods have focused on dichotomous phenotypes such as case/control status or quantitative phenotypes such as weight or cholesterol level. Rare variant association methods for multinomial phenotypes, or categorical outcomes with more than two possibilities, have not been adequately addressed. The purpose of this study is to develop new methods of rare variant association analysis for a multinomial phenotype. Several new methods are proposed and evaluated using simulations.

Findings and Conclusions:

Simulations showed that two of the proposed methods are viable for rare variant association analysis with multinomial phenotypes. These methods have the correct or conservative Type I error rate and reasonable power for large samples with a moderate heritability. The viable methods are applied to resequencing data from the Dallas Heart Study. One of the methods detected an association between a categorized plasma triglyceride level and the ANGPTL3 and ANGPTL4 genes.