Testing pleiotropy vs. separate QTL in multiparental populations

Abstract

8The high mapping resolution of multiparental populations, combined with technology to measure tens of thousands of phenotypes, presents a need for quantitative methods to enhance understanding of the genetic architecture of complex traits. When multiple traits map to a common genomic region, knowledge of the number of distinct loci provides important insight into the underlying mechanism and can assist planning for subsequent experiments. We extend the method of Jiang and Zeng (1995), for testing pleiotropy with a pair of traits, to the case of more than two alleles. We also incorporate polygenic random effects to account for population structure. We use a parametric bootstrap to determine statistical significance. We apply our methods to a behavioral genetics data set from Diversity Outbred mice. Our methods have been incorporated into the R package qtl2pleio. 9 10 11 12 13 14 15 16 17 KEYWORDS Quantitative trait locus; pleiotropy; multivariate analysis; linear mixed effects models; systems genetics 18 19 20Complex trait studies in multiparental populations present new challenges in statistical methods 21 and data analysis. Among these is the development of strategies for multivariate trait analysis. The 22 joint analysis of two or more traits allows one to address additional questions, such as whether two 23 traits share a single pleiotropic locus. 24Previous research addressed the question of pleiotropy vs. separate QTL in two-parent crosses. 25Jiang and Zeng (1995) developed a likelihood ratio test for pleiotropy vs. separate QTL for a pair 26 of traits. Their approach assumed that each trait was affected by a single QTL. Under the null 27 hypothesis, the two traits were affected by a common QTL, and under the alternative hypothesis the 28 two traits were affected by distinct QTL. Knott and Haley (2000) used linear regression to develop 29 a fast approximation to the test of Jiang and Zeng (1995), while Tian et al. (2016) used the methods 30 from Knott and Haley (2000) to dissect QTL hotspots in a F 2 population. 31 Multiparental populations, such as the Diversity Outbred (DO) mouse population (Churchill et al. 32 2012), enable high-precision mapping of complex traits (de Koning and McIntyre 2014). The DO 33 mouse population began with progenitors of the Collaborative Cross (CC) mice (Churchill et al. 34 2004) Each DO mouse is a highly heterozygous genetic mosaic of alleles from the eight CC founder 35 lines. Random matings among non-siblings have maintained the DO population for more than 23 36 generations (Chesler et al. 2016). 37 Several limitations of previous pleiotropy vs. separate QTL tests prevent their direct application 38 in multiparental populations. First, multiparental populations can have complex patterns of relat-39 edness among subjects, and failure to account for these patterns of relatedness may lead to spurious 40 results (Yang et al. 2014). Second, previous tests allowed for only two founder lines (We developed a pleiotropy vs. separate QTL test for two traits in multiparental populatio...

Host-pathogen genetic interactions underlie tuberculosis susceptibility

“…In this interval, the CAST haplotype was associated with both low bacterial burden and CXCL1 ( Figure 2F). At both Tip5 and Tip8, we found no statistical evidence that the positions of the associated QTL were different (Tip5 p=0.55; Tip8 p=0.27; bootstrap samples) (Boehm et al, 2019). These observations support the role of a single causal variant at each locus that is responsible for a pleiotropic trait.…”

“…Both the relative positions of the module-associated QTL and the associated founder haplotypes indicated that a single genetic variant controlled the abundance of ESX1 and mbt mutants (Hip42). Specifically, we found no statistical support for differentiating these QTL based on position (p=0.93) (Boehm et al, 2019) and the same founder haplotypes were associated with extreme trait values at both loci, though they had opposite effects on the abundance of ESX1 mutants and mbt mutants ( Figure 6F). We conclude that a single haplotype has a pleiotropic effect on Mtb's environment and has opposing effects of the requirement for mycobactin synthesis and ESX1 secretion.…”

“…LOD profiles and effect plots were generated using the plotting functions of the R/qtl2 package. Multiple QTL at similar genetic locations were assessed for independence using qtl2pleio with 400 bootstrap samples (Boehm et al, 2019). The quantile-based permutation thresholding method of Neto et al (Neto et al, 2012) was used to assess the statistical significance of co-mapping traits.…”

Covariation of brain and skull shapes as a model to understand the role of crosstalk in development and evolution

“…The test statistic is calculated as the logarithm of the ratio between these two likelihood values. Significance testing of pleiotropy versus separate QTL was obtained via parametric bootstrapping (Boehm et al, 2019). A significant result indicates the two traits localize to a distinct QTL (i.e., p < .05), while a nonsignificant result indicates the two traits localize to a common QTL (i.e., p > .05).…”