Association in Multifactorial Traits: How to Deal with Rare Observations?

Abstract

To detect the role of a candidate gene for a trait in a sample of individuals, we may test SNP haplotype or diplotype effects. For a limited sample size, many haplotype or diplotype categories may contain few individuals. This involves a power decrease when testing the association between the trait and the haplotypes or diplotypes as these categories provide little additional information while increasing the degrees of freedom. The present paper proposes a new strategy to group rare categories based on a measure of similarity between haplotypes or diplotypes and compares it to two other possible strategies to deal with rare categories: a SNP selection strategy based on haplotype diversity, and a grouping strategy that pools all rare categories into a single baseline group. This comparison is performed by means of simulation under four scenarios. We show that this new strategy shows the largest increase in power irrespective of the model underlying the candidate gene in the studied trait. This strategy therefore provides a powerful alternative to currently used methods to reduce the number of rare categories.

Statistical performance of cladistic strategies for haplotype grouping in pharmacogenetics

“…Tables IV and V show the power and Type I error results for the test of genetic effects overall and the test of treatment by clade interaction effects, respectively. It is generally recognized that the analyst must decide how to eliminate the rare haplotype categories prior to executing the analysis [3,4] and it has been shown that removing rare haplotype categories from the analysis leads to greater power [10][11][12], although it is not always clear what level of haplotype diversity should be retained (see Reference [12] for a strategy based on Shannon information content). As shown in Tables IV and V, our findings also indicated that power was usually greater when rare categories were removed prior to testing.…”

“…As an example, Figure 2 shows the entire cladogram for the 15 PPAR haplotypes under the haplotype frequencies used for simulation. For those simulations that eliminated any model terms for rare haplotypes prior to testing, each rare haplotype (frequency <0.05) was grouped with the nearest (in mutational steps) common haplotype, always equating to the more frequent common haplotype in the event of equal mutational distances (similar to the method in Reference [10]). The cladogram was then constructed from the common haplotypes using the frequency-ordered version of Prim's algorithm.…”

Estimating Haplotype Effects on Dichotomous Outcome for Unphased Genotype Data Using a Weighted Penalized Log-Likelihood Approach

“…Other approaches to deal with rare haplotypes, like pooling them into one category, or pooling them with common haplotypes that are very similar, lead to pooled categories that are hard to interpret. These methods seem to increase power [20] , but only in specific situations where pooled haplotypes have similar effects.…”

On the use of phylogeny‐based tests to detect association between quantitative traits and haplotypes

“…A first set of approaches focus on rare haplotypes that can either be grouped together [Tzeng et al, 2003;Jannot et al, 2004] or with the most similar common haplotype [Jannot et al, 2004;Tzeng, 2005;Tzeng et al, 2006]. In a second set of approaches, all haplotypes are grouped together according to various clustering algorithms: principal components [Sha et al, 2005], classification trees [Yu et al, 2005], distance methods [Durrant et al, 2004], clustering based on the sharing of haplotype segments [Li et al, 2006] or evolutionary-based clustering [Templeton et al, 1987Seltman et al, 2001Seltman et al, , 2003Bardel et al, 2005].…”