Investigating the fluctuations of residual polygenic variance on predictive ability of genomic breeding values of crossbred progeny

Document Type : Research Paper

Authors

• Somayeh Barani ¹
• S Reza Miraei-Ashtiani ¹
• Ardeshir Nejati-Javaremi ²
• Hadi Esfandyari ³
• Majid Khansefid ⁴

¹ Department of Animal Science, College of Agriculture and Natural Resources, University of Tehran, Karaj, Iran

² Department of Animal Science, College of Agriculture and Natural Resources, university of Tehran, Karaj, Iran

³ Norwegian Beef Cattle Organizations, TYR, Hamar, Norway

⁴ Agriculture Victoria Research, Bundoora, VIC 3083, Australia School of Applied Systems Biology, La Trobe University, Bundoora, VIC 3083, Australia

Abstract

Keywords

• Crossbreeding
• Single step genomic evaluation
• Prediction ability
• Simulation

Main Subjects

• Animal Breeding and Genetics

Extended Abstract

Introduction

Crossbreeding, as outcrossing between breeds, is widely used in most livestock breeds to benefit from hybrid vigor effects and breed complementarity to improve performance levels of crossbred offspring. Proper crossbreeding plans can also prevent inbreeding and inbreeding depression. In fact, the main reasons to use crossbreeding plans are: 1- Using different levels of additive genetic effects between breeds, 2- Take advantage of hybrid vigor due to non-additive effects such as epistasis, dominance and over dominance and also, prevent inbreeding and inbreeding depression. However, collecting information on crossbred populations is difficult and expensive so, the breeding value of crossbreds is inevitably predicted based on the information of the parental purebreds. The genomic evaluation of crossbred offspring in order to increase their performance, is primarily based on the information of purebred parents, while it eventually leads to the selection of superior purebreds in the parental lines (Esfandyari et al., 2015). Nevertheless, A variety of factors such as genetics-environment interactions, linkage disequilibrium, non-additive genetic effects such as dominance, epistasis and imprinting, reduce the genetic correlation between these two populations, therefore the selection based on purebred parental populations will not necessarily produce crossbred offspring with higher performance. However, due to the low genetic correlation between the two populations, purebred animals do not always produce crossbred offspring with enough high performance. In general, predicting the breeding value of each population requires information from the reference population of the same population, but inevitably, genetic evaluation of crossbred uses information from mixed populations with high genetic correlation.

Additionally, Single-step Genomic Best Liner Unbiased Prediction (ssGBLUP) models can simultaneously use genotyped(G) and non-genotyped animal’s information () which leads to realize more accurate genomic breeding values.  But then, one of the major challenges in using ssGBLUP and simultaneously using pedigree and genomic matrices of genotyped and non-genotyped animals is compromised with incompatibility between the two matrices, which can lead to bias and dispersion of genomic estimated breeding values. The compatibility of these two matrices means that the mean of diagonal and non-diagonal elements of G and  must be similar. To calculate , is blended to a small part of a positive definite matrix, which usually includes either an Identity matrix or , but the condition for blending  with would be inversable and compatible (Vitezica et al., 2011). One simple way to reduce the bias and decrease the predicted inflation and deflation of GEBV is to consider inbreeding coefficient in calculation of inverse of the relationship matrices and reliability, besides using the appropriate proportion of residual polygenic variance in blending of genomic and pedigree relationship matrices. So, the use of multiple populations and appropriate genomic evaluation models to predict of crossbred performance has been a subject of interest of this investigation.

Material methods

In this study, the datasets of purebred and crossbred animals were simulated based on a sheep production system for a single trait with a heritability of 0.3. This simulation was implemented with 50K single nucleotide polymorphisms and 500 quantitative trait loci (QTLs) across the whole genome. Minor allele frequencies of markers were assumed upper than 0.05 and mutation rates of the SNPs and QTL were considered. In this situation, the QTL allele effects are inferred from a Gamma distribution allowing for.  True breeding value (TBV) has counted the sum of the additive effects for each QTL and the observed phenotypes were computed additive effect and residual effects. The ability to predict GEBV in crossbreds was investigated using ssGBLUP method and simulated information of purebred parental and crossbred progenies. Further, we investigated the effect of proportion of residual polygenic variance in blending of genomic and pedigree relationship matrices in order to improve the prediction ability.

Results

 The results showed that using genotype of crossbred animals in combination with purebred parental population in genomic evaluation improved the prediction accuracy, when the breeding objective was to increase the crossbreed performance. In addition, considering inbreeding coefficient in calculation of the inverse of the relationship matrix and the consistency of prediction decreased the prediction bias and increased the reliability of GEBV. Furthermore, different residual polygenic variance (β) values did not have a significant effect on the accuracy, bias and inflation or deflation of the prediction of GEBV in crossbred animals.

Conclusion

 The prediction accuracy, bias, and dispersion were similar across the different proportion of residual polygenic variance (β) in blending the genomic and pedigree relationship matrices, therefore, for the simplicity of ssGBLUP model, it is recommended to use the default β (0.05).