Identifying causal rare variants of disease through family-based analysis of Genetics Analysis Workshop 17 data set

Abstract

Linkage- and association-based methods have been proposed for mapping disease-causing rare variants. Based on the family information provided in the Genetic Analysis Workshop 17 data set, we formulate a two-pronged approach that combines both methods. Using the identity-by-descent information provided for eight extended pedigrees (n = 697) and the simulated quantitative trait Q1, we explore various traditional nonparametric linkage analysis methods; the best result is obtained by assuming between-family heterogeneity and applying the Haseman-Elston regression to each pedigree separately. We discover strong signals from two genes in two different families and weaker signals for a third gene from two other families. As an exploratory approach, we apply an association test based on a modified family-based association test statistic to all rare variants (frequency < 1% or < 3%) designated as causal for Q1. Family-based association tests correctly identified causal single-nucleotide polymorphisms for four genes (KDR, VEGFA, VEGFC, and FLT1). Our results suggest that both linkage and association tests with families show promise for identifying rare variants.

Figure 1

Detection rates from modified FBAT for all genes on chromosomes 4, 5, 6, and 13. Each bar in the graphs represents the percentage of times that the gene was significant (p < 0.05) in the 200 replicates. True-positive disease genes are labeled. Of note, the KIT locus on chromosome 4, frequently detected as a false positive, is in close proximity (394 kb) to the disease-causing KDR locus.