Estimation & Prediction

Validation of Indirect Predictions for Traits With Low to Moderate Heritabilities

---

Abstract

Validation of indirect genomic predictions for traits with low to moderate heritabilities John Thomason, Fernando Bussiman, Zuleica Trujano, Sergio Sanchez, Andre Lima, Andre Garcia, Daniela Lourenco Indirect predictions (IP) are a valuable tool for modern breeding programs that have an ever-increasing number of genotyped animals. It provides genomic predictions based on SNP effects and SNP content between official evaluation runs, for young selection candidates and commercial animals that may never make it into an official evaluation. However, the validity of IP depends on its accuracy, and there is evidence that the method fails to provide accurate predictions for traits with low heritability. Our objectives in this study were to investigate whether IP are reliable for traits with low heritability, and whether the choice of core animals in the Algorithm for Proven and Young (APY) can impact IP. Data were provided by the American Angus Association (AAA), which uses IP mainly as interim evaluations and through the GeneMax advantage program, where they provide genomic scores for commercial heifers. IP were computed for 19 traits with heritabilities varying from 0.04 to 0.56, using the blupf90 suite of programs. Available data included over 15 million phenotypic records and 12 million animals in the pedigree, of which 1.7 million were genotyped. First, benchmark GEBV were calculated with all the available data, including genotypes from validation animals, via single-step GBLUP (ssGBLUP) using APY. Then, data for the validation animals were removed; thus, GEBV were computed based on information for training animals (all but the validation animals), and back-solved to obtain SNP effects. Following, the IP for validation animals were computed based on SNP effects and SNP content. The mean GEBV from the main evaluation (excluding animals having IP) was added to IP, so that IP has similar mean to the benchmark GEBV. Finally, IP for the validation animals were compared to their benchmark GEBV. For traits with high heritability and millions of records, GEBV and IP were nearly identical, with Pearson and Spearman correlations greater than 0.99. When examining lower heritability traits with less data, correlations remained greater than 0.99, except for some traits evaluated using random regression models, which have a correlation of 0.97. These results followed similar trends independently of the choice of core animals in APY. Although slight differences exist between GEBV and IP for traits with low heritability, these are negligible, which demonstrates that IP can be used as interim predictions across traits with varying heritabilities.

Keywords: 2026

How to Cite:

Thomason, J., Bussiman, F., Lima, A., Sanchez, S., Trujano, Z., Garcia, A. & Lourenco, D., (2026) “Validation of Indirect Predictions for Traits With Low to Moderate Heritabilities”, World Congress on Genetics Applied to Livestock Production Digital Archive 2026(1): 2257193. doi: https://doi.org/10.31274/wcgalp.23388

Rights: 1