Logo image
Back
Markers as traits in multivariate BLUP: using REML for association testing and integration with breeding value prediction
Abstract   Peer reviewed

Markers as traits in multivariate BLUP: using REML for association testing and integration with breeding value prediction

Jeremy T Brawner and M J Dieters
4th International Conference on Quantitative Genetics Programme and Book of Abstracts, p.60
International Conference on Quantitative Genetics (ICQG): Understanding Variation in Complex Traits, 4th (Edinburgh, United Kingdom, 17-Jun-2012–22-Jun-2012)
2012
url
http://icqg2012.org.uk/View
Webpage

Abstract

Forestry Sciences Plant Biology
A pedigreed population of an outcrossing forest tree species that was established in multiple progeny trials and genotyped for an association study is used to demonstrate the concept of using markers as additional traits in a multivariate analysis. The population exhibited moderate to high narrow-sense heritability estimates for three economically important traits: stem diameter (0.28), wood density (0.44) and pulp yield and (0.36). For the 65 markers that were sequenced within 12 candidate genes, P-values were estimated for four tests of association using two different approaches: 1) two software packages (Tassle and Asreml) fit standard univariate mixed models that include each SNP as an additional fixed effect, and 2) bivariate or multivariate mixed models including each SNP as additional selection trait. The first approach determines if a significant proportion of phenotypic variance is attributable to the marker effect using an F-test and the second estimates the significance of the genetic correlation between additive genetic values and marker frequency using log-likelihood ratio tests. Within either the univariate or multivariate approach there was a significant correlation between the tests of significance. The two programs used for the first approach produced nearly identical results and the significance of marker-trait correlations generated with the bivariate and multivariate analyses were equally similar. However, there was little correspondence between the two approaches: only one of the 12 among-approach correlations was weakly positive (r=0.25 and p<0.05). This empirical study demonstrates that different SNP markers would be selected using multivariate analytical methods compared to the standard marker:trait association methods. In addition to demonstrating empirical results, the presentation provides an interpretation of genetic parameters estimated using REML derived variance components within a multivariate framework including two phenotypes and two markers as traits, including: 1) typical heritability of phenotypic traits, 2) heritability of genic traits, 3) typical trait:trait correlations, 4) maker:trait correlations as association tests, and 5) marker:marker correlations as LD estimates.

Details

Metrics

227 Record Views
Logo image