QTL detection in multi-breed cattle populations including crossbred animals: a simulation study

Abstract

Many methods have been proposed to identify regions neighbouring causal variants, primarily targeting purebred populations, and modified to also handle multi-breed and crossbred populations. The power of such methods is affected by differences in linkage disequilibrium phase between markers and quantitative trait loci (QTL) across breeds. In this study, we aimed to provide a novel method based on the breed origin of alleles, referred to as breed origin of alleles (BOA), which allows QTL effects to be different but correlated across breeds and suitable for analysing purebred (PB) and crossbred (XB) animals' data together (PB+XB_BOA). Using simulations, we compared the results from our method with those obtained with a conventional BayesBπ when the pure breeds were analysed individually (PB_SINGLE) and when all the pure breeds and crossbred animals' data were simply combined without accounting for BOA (PB+XB_JOINT). The simulated data comprised a multi-breed population of which 65% were purebreds, and 35% were crossbreds, with a varied contribution of the three pure breeds. True positives, power of detection, false positive rate, positive predictive value and score of each model were used for comparison. The analysis on PB_SINGLE presented the lowest ability to detect QTL, especially in the smallest breed. When analysing the animals jointly, more QTL were identified in all breeds, but even more were detected when accounting for BOA, which in addition, led to a higher power of detection and positive predictive value while keeping narrower peaks than ignoring it. We concluded that the methodology proposed outperformed the other two approaches in identifying genomic regions of interest, one analysing purebred data separately and the other combining purebred and crossbred data in a single analysis.

Keywords: Across-breeds; Admixed populations; Bayesian analysis; Breed-origin-of-alleles; Genomic regions.