Temporal and genomic analysis of additive genetic variance in breeding programmes

Abstract

This study demonstrates a framework for temporal and genomic analysis of additive genetic variance in a breeding programme. Traditionally we used specific experimental designs to estimate genetic variance for a specific group of individuals and a general pedigree-based model to estimate genetic variance for pedigree founders. However, with the pedigree-based model we can also analyse temporal changes in genetic variance by summarising sampled realisations of genetic values from a fitted model. Here we extend this analysis to a marker-based model and build a framework for temporal and genomic analyses of genetic variance. The framework involves three steps: (i) fitting a marker-based model to data, (ii) sampling realisations of marker effects from the fitted model and for each sample calculating realisations of genetic values, and (iii) calculating variance of the sampled genetic values by time and genome partitions. Genome partitions enable estimation of contributions from chromosomes and chromosome pairs and genic and linkage-disequilibrium variances. We demonstrate the framework by analysing data from a simulated breeding programme involving a complex trait with additive gene action. We use the full Bayesian and empirical Bayesian approaches to account for the uncertainty due to model fitting. We also evaluate the use of principal component approximation. Results show good concordance between the simulated and estimated variances for temporal and genomic analyses and give insight into genetic processes. For example, we observe reduction of genic variance due to selection and drift and buildup of negative linkage-disequilibrium (the Bulmer effect) due to directional selection. In this study the popular empirical Bayesian approach estimated the variances well but it underestimated uncertainty of the estimates. The principal components approximation biases estimates, in particular for the genic variance. This study gives breeders a framework to analyse genetic variance and its components in different stages of a programme and over time.

Estimation of the autosomal contribution to total additive genetic variability of carcass traits in Japanese Black cattle

“…In fact, continued selection in Japanese Black population may have created LD across chromosomes (Bulmer, 1971). Based on the approach shown by Lehermeier et al (2017), Lara et al (2022) assessed the performance of partitioning the total additive genetic variance into genic variance, covariance due to intra‐chromosomal LD, and that due to inter‐chromosomal LD by using estimated marker effects, a different approach from ours. A more sophisticated way to cope with LD of two SNPs on different chromosomes might be to remove one SNP from the analysis: that is, removing the SNPs on the remaining chromosomes which are in LD with the SNPs on a particular chromosome in G 1 .…”

Impact of Genomic Prediction Model, Selection Intensity, and Breeding Strategy on the Long-Term Genetic Gain and Genetic Erosion in Soybean Breeding

“…In soybeans, the management of diversity is necessary to ensure useful variability for future breeding objectives, such as yield performance under drought or waterlogging (Valliyodan et al, 2017), the seed oil and protein content profiles (Stewart-Brown et al, 2019), and disease resistance (de Azevedo Peixoto et al, 2017). Monitoring genetic diversity in the genomic era can be performed through tracking overtime changes in allele frequencies (Allier et al, 2019b;de Castro Lara et al, 2020;Meuwissen et al, 2020). We showed that selection could quickly exhaust genetic diversity under closed breeding systems, and breeding systems can benefit from balancing short gains to preserve diversity and assure long-term gains.…”

SimpleMating: R-package for Prediction and Optimization of Breeding Crosses Using Genomic Selection

“…The genetic standard deviation was estimated as the standard deviation of the standardized genetic values. The genic standard deviation was estimated as the square root of the variance of genetic values assuming no linkage between the causal loci, directly obtained from AlphaSimR parental's population Lara et al, 2022;Pocrnic et al, 2023). All scripts for the implementation of the simulated scenarios and to plot and analyze the results can be found at: https://github.com/Resende-Lab/MaizeOCS-SimpleMating.…”