Impact of different genomic relationship matrix construction methods on the accuracy of genomic prediction in different species

Abstract

Objective: Genomic best linear unbiased prediction (GBLUP) is a key method in genomic prediction, relying on the construction of a genomic relationship matrix (G-matrix). Although various methods for G-matrix construction have been proposed, the performance of these methods across different species has not been thoroughly compared.

Methods: This study systematically evaluated the performance of six genomic relationship matrix (G-matrix) construction methods in improving the prediction accuracy of GBLUP models across four species: pigs, bulls, wheat, and mice. The methodological framework included: (1) an initial unscaled matrix; (2) five scaled methods utilizing allele frequency centralization. The scaled methods comprised: (a) three variance-weighted approaches using allele frequencies fixed at 0.5 (G05), observed frequencies (GOF), or average minor allele frequencies (GMF); (b) two centralized methods with weighting by either the trace of the numerator matrix (GN) or reciprocals of each locus's expected variance (GD).

Results: The GD matrix demonstrated significant prediction accuracy improvements for pig traits. Conversely, most scaled G-matrices showed minimal effects on mice, wheat, and bull, even with underperforming unscaled baselines in prediction accuracy compared to the original unscaled matrix. The learning curve for bull data showed the choice of G-matrix had minimal impact on prediction accuracy when the reference population size and genetic marker density reached a certain threshold.

Discussion: The study concluded that the optimal G-matrix construction method varies across species, with population structure being a key factor. These findings highlight the importance of species-specific optimization in genomic prediction and suggest that the influence of G-matrix construction diminishes in large-scale, high-density genomic datasets.

Keywords: accuracy of prediction; different species; genomic relationship matrix; marker density; size of reference population.

FIGURE 1

Coefficient of determination ( $R^2$ ) using different G-matrices in pig data. bf 10, 10^th rib backfat of the pig; lastrib, last rib backfat of the pig; LMA, loin muscle area at the 10^th rib of the pig.

FIGURE 2

Coefficient of determination ( $R^2$ ) using different G-matrices in mice data. Obesity_BMI, body mass index; EndNormalBW, body weight; Obesity_BodyLength, body length. Obesity_BMI and Obesity_BodyLength both had low values of $R^2$ , less than 0.2. The value of $R^2$ of Obesity_BW is slightly higher than that of the other two but not more than 0.5. In general, the accuracy of the GBLUP model in predicting Obesity_BMI, Obesity_BL, and Obesity_BW of mice was low. Values of $R^2$ of Obesity_BL predicted by six types of G-matrices are almost the same, which is approximately 0.17. For values of $R^2$ predicted by Obesity_BW and Obesity_BMI, G05 and GD were significantly higher than MM′. GOF and GN obtained similar accuracy for all these three traits.

FIGURE 3

Coefficient of determination ( $R^2$ ) using different G-matrices in the wheat data. E means environment.

FIGURE 4

Coefficient of determination ( $R^2$ ) using different G-matrices in the bull data. FP, milk fat percentage; MY, milk yield; SCS, somatic cell score.

FIGURE 5

(A) Summary of the number of individuals and genetic markers. (B) Comparison of average values of $R^2$ .

FIGURE 6

Effects of reference population size and marker density on the accuracy of genomic prediction across different G-matrices in bull data. (A) Coefficient of determination for milk yield using different G-matrices and the number of individuals. (B) Coefficient of determination for milk fat percentage using different G-matrices and the number of individuals. (C) Coefficient of determination for SCS using different G-matrices and the number of individuals. (D) Coefficient of determination for milk yield using different G-matrices and the number of markers.