Genetic heterogeneity and trans regulators of gene expression

Abstract

Heterogeneity poses a challenge to linkage mapping. Here, we apply a latent class extension of Haseman-Elston regression to expression phenotypes with significant evidence of linkage to trans regulators in 14 large pedigrees. We test for linkage, accounting for heterogeneity, and classify individual families as "linked" and "unlinked" on the basis of their contribution to the overall evidence of linkage.

Summary of Genetic Analysis Workshop 15: Group 9 linkage analysis of the CEPH expression data

“…These methods included (1) explicitly modeling heterogeneity with a liability class model that allowed for linked and unlinked families in HE linkage analysis and based estimation of significance on a permutation approach [Bastone et al, 2007]; (2) modeling heterogeneity with a standard model that maximizes the likelihood over a heterogeneity parameter [Christensen et al, 2007]; and (3) simply presenting results on individual families for parametric analysis [Sung et al, 2007a], Merlin-regress analysis [Huang et al, 2007], and HE analysis [Xing et al, 2007], respectively. All five groups came to the same conclusion that for many traits, there was evidence for linkage in only a small number of families -often only two to three of the 14 available families.…”

“…Several groups found coincident linkage signals for different traits [Bastone et al, 2007;Diao and Lin, 2007;Franceschini et al, 2007;Lantieri et al, 2007;Papachristou et al, 2007;Wang et al, 2007]. While some interpreted this as evidence for coordinated regulation, there is need for caution in this interpretation since the resolution in the data set could not discriminate among the existence of multiple eQTLs in a region, the presence of a regulator locus that affects multiple traits, and chance correlation introduced by a large number of traits measured on a modest number of subjects.…”

“…Two groups estimated more explicit trait models for a subset of traits and used these models for parametric linkage analyses: Sung et al [2007a] estimated QTL parameters with Bayesian oligogenic segregation analysis and collapsed multiple QTLs into one to two QTLs with or without additional polygenic variance, and Christensen et al [2007] based their models on dominant and recessive models from a previous publication [Smith et al, 1996] and used these to compute heterogeneity LOD scores (HLODs). Finally Bastone et al [2007] developed a latent class extension to the Haseman-Elston (HE) regression approach [Haseman and Elston, 1972], to incorporate locus heterogeneity into the linkage analysis and to estimate the number of families with evidence for segregation of an eQTL in a particular region.…”

A Latent Class Model for Testing for Linkage and Classifying Families when the Sample May Contain Segregating and Non-Segregating Families

“…The procedure yields valid type I error rates, and allows the procedure to be used irrespective of the bias in the estimate, and whether a two-class model can be fit to the observed data. In developing this approach, we initially used a permutation procedure where we accepted the null hypothesis when the two-class model failed to converge [13,26] . In simulation studies, this approach yielded acceptable type I error rates, but, obviously, failed to detect linkage in any gene where the two-class model did not fit and yielded lower power than the procedure detailed here.…”

“…We briefly introduced a latent class approach to this problem in an earlier study of trans regulators of gene expression [13] . Here we describe in detail an improvement on the original testing approach using examples of cis regulation of gene expression as motivation.…”