Application of linear and machine learning models to genomic prediction of fatty acid composition in Japanese Black cattle.

We collected 3180 records of oleic acid (C18:1) and monounsaturated fatty acid (MUFA) measured using gas chromatography (GC) and 6960 records of C18:1 and MUFA measured using near-infrared spectroscopy (NIRS) in intermuscular fat samples of Japanese Black cattle. We compared genomic prediction performance for four linear models (genomic best linear unbiased prediction [GBLUP], kinship-adjusted multiple loci [KAML], BayesC, and BayesLASSO) and five machine learning models (Gaussian kernel [GK], deep kernel [DK], random forest [RF], extreme gradient boost [XGB], and convolutional neural network [CNN]). For GC-based C18:1 and MUFA, KAML showed the highest accuracies, followed by BayesC, XGB, DK, GK, and BayesLASSO, with more than 6% gain of accuracy by KAML over GBLUP. Meanwhile, DK had the highest prediction accuracy for NIRS-based C18:1 and MUFA, but the difference in accuracies between DK and KAML was slight. For all traits, accuracies of RF and CNN were lower than those of GBLUP. The KAML extends GBLUP methods, of which marker effects are weighted, and involves only additive genetic effects; whereas machine learning methods capture non-additive genetic effects. Thus, KAML is the most suitable method for breeding of fatty acid composition in Japanese Black cattle.

Exploring the size of reference population for expected accuracy of genomic prediction using simulated and real data in Japanese Black cattle.

BACKGROUND: Size of reference population is a crucial factor affecting the accuracy of prediction of the genomic estimated breeding value (GEBV). There are few studies in beef cattle that have compared accuracies achieved using real data to that achieved with simulated data and deterministic predictions. Thus, extent to which traits of interest affect accuracy of genomic prediction in Japanese Black cattle remains obscure. This study aimed to explore the size of reference population for expected accuracy of genomic prediction for simulated and carcass traits in Japanese Black cattle using a large amount of samples. RESULTS: A simulation analysis showed that heritability and size of reference population substantially impacted the accuracy of GEBV, whereas the number of quantitative trait loci did not. The estimated numbers of independent chromosome segments (Me) and the related weighting factor (w) derived from simulation results and a maximum likelihood (ML) approach were 1900-3900 and 1, respectively. The expected accuracy for trait with heritability of 0.1-0.5 fitted well with empirical values when the reference population comprised > 5000 animals. The heritability for carcass traits was estimated to be 0.29-0.41 and the accuracy of GEBVs was relatively consistent with simulation results. When the reference population comprised 7000-11,000 animals, the accuracy of GEBV for carcass traits can range 0.73-0.79, which is comparable to estimated breeding value obtained in the progeny test. CONCLUSION: Our simulation analysis demonstrated that the expected accuracy of GEBV for a polygenic trait with low-to-moderate heritability could be practical in Japanese Black cattle population. For carcass traits, a total of 7000-11,000 animals can be a sufficient size of reference population for genomic prediction.