Genome-Enabled Prediction Methods Based on Machine Learning

Growth of artificial intelligence and machine learning (ML) methodology has been explosive in recent years. In this class of procedures, computers get knowledge from sets of experiences and provide forecasts or classification. In genome-wide based prediction (GWP), many ML studies have been carried out. This chapter provides a description of main semiparametric and nonparametric algorithms used in GWP in animals and plants. Thirty-four ML comparative studies conducted in the last decade were used to develop a meta-analysis through a Thurstonian model, to evaluate algorithms with the best predictive qualities. It was found that some kernel, Bayesian, and ensemble methods displayed greater robustness and predictive ability. However, the type of study and data distribution must be considered in order to choose the most appropriate model for a given problem.

Benefits of testing in both bio-secure and production environments in genomic selection breeding programs for commercial broiler chicken

Background

A breeding program for commercial broiler chicken that is carried out under strict biosecure conditions can show reduced genetic gain due to genotype by environment interactions (G × E) between bio-secure (B) and commercial production (C) environments. Accuracy of phenotype-based best linear unbiased prediction of breeding values of selection candidates using sib-testing in C is low. Genomic prediction based on dense genetic markers may improve accuracy of selection. Stochastic simulation was used to explore the benefits of genomic selection in breeding schemes for broiler chicken that include birds in both B and C for assessment of phenotype.

Results

When genetic correlations (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$r_{g}$$\end{document}rg) between traits measured in B and C were equal to 0.5 and 0.7, breeding schemes with 15, 30 and 45% of birds assessed in C resulted in higher genetic gain for performance in C compared to those without birds in C. The optimal proportion of birds phenotyped in C for genetic gain was 30%. When the proportion of birds in C was optimal and genotyping effort was limited, allocating 30% of the genotyping effort to birds in C was also the optimal genotyping strategy for genetic gain. When \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$r_{g}$$\end{document}rg was equal to 0.9, genetic gain for performance in C was not improved with birds in C compared to schemes without birds in C. Increasing the heritability of traits assessed in C increased genetic gain significantly. Rates of inbreeding decreased when the proportion of birds in C increased because of a lower selection intensity among birds retained in B and a reduction in the probability of co-selecting close relatives.

Conclusions

If G × E interactions (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$r_{g}$$\end{document}rg of 0.5 and 0.7) are strong, a genomic selection scheme in which 30% of the birds hatched are phenotyped in C has larger genetic gain for performance in C compared to phenotyping all birds in B. Rates of inbreeding decreased as the proportion of birds moved to C increased from 15 to 45%.

Genetic parameters of foot-pad dermatitis and body weight in purebred broiler lines in 2 contrasting environments

The aims of this study were to investigate the genetic background of foot-pad dermatitis (FPD) in 4 different broiler lines reared in 2 contrasting environments (pedigree or sib-test) and to evaluate the performance of simultaneous genetic selection for improved FPD and BW. Data were available for 4 generations from 4 broiler lines, bred with varying intensities of selection for growth. The average BW ranged from 1.7 to 2.4 kg at 5 wk of age. In the pedigree environment, the prevalence of FPD ranged from 14 to 37%, with 3 to 9% being severely affected; in the sib-test environment, these values were correspondingly 45 to 79% and 35 to 70%. Both traits showed re-ranking of the 4 lines in terms of phenotype across the 2 environments, indicating the existence of a genotype-by-environment interaction. In both environments, females showed higher prevalences of FPD than males. In line with their higher prevalence, heritabilities of FPD in the sib-test environment ranged from 0.22 to 0.32, compared with 0.18 to 0.24 for FPD in the pedigree environment (all SE ≤0.02). Estimates of the genetic correlation between FPD in the pedigree and in the sib-test environments were high (0.78-0.82), which suggests that selection against FPD in a highly biosecure environment can improve the genetic merit for birds reared under commercial conditions. Estimates of the genetic associations between FPD and BW were small and varied in sign. Predicted responses to selection showed a yearly reduction in average score of -3.4 to -7.5% for FPD in the pedigree environment and -0.5 to -6.6% for FPD in the sib-test environment, while maintaining improvement of BW of 2.6 to 3.2% and 2.6 to 3.8% of the average BW per year, respectively. This research indicates that balanced genetic selection for both BW and FPD in contrasting environments is an effective strategy to reduce the genetic disposition to develop FPD in broilers.

Comparison of classification methods for detecting associations between SNPs and chick mortality

Multi-category classification methods were used to detect SNP-mortality associations in broilers. The objective was to select a subset of whole genome SNPs associated with chick mortality. This was done by categorizing mortality rates and using a filter-wrapper feature selection procedure in each of the classification methods evaluated. Different numbers of categories (2, 3, 4, 5 and 10) and three classification algorithms (naïve Bayes classifiers, Bayesian networks and neural networks) were compared, using early and late chick mortality rates in low and high hygiene environments. Evaluation of SNPs selected by each classification method was done by predicted residual sum of squares and a significance test-related metric. A naïve Bayes classifier, coupled with discretization into two or three categories generated the SNP subset with greatest predictive ability. Further, an alternative categorization scheme, which used only two extreme portions of the empirical distribution of mortality rates, was considered. This scheme selected SNPs with greater predictive ability than those chosen by the methods described previously. Use of extreme samples seems to enhance the ability of feature selection procedures to select influential SNPs in genetic association studies.

Inferring genetic values for quantitative traits non-parametrically

Inferences about genetic values and prediction of phenotypes for a quantitative trait in the presence of complex forms of gene action, issues of importance in animal and plant breeding, and in evolutionary quantitative genetics, are discussed. Current methods for dealing with epistatic variability via variance component models are reviewed. Problems posed by cryptic, non-linear, forms of epistasis are identified and discussed. Alternative statistical procedures are suggested. Non-parametric definitions of additive effects (breeding values), with and without employing molecular information, are proposed, and it is shown how these can be inferred using reproducing kernel Hilbert spaces regression. Two stylized examples are presented to demonstrate the methods numerically. The first example falls in the domain of the infinitesimal model of quantitative genetics, with additive and dominance effects inferred both parametrically and non-parametrically. The second example tackles a non-linear genetic system with two loci, and the predictive ability of several models is evaluated.