Estimation & Prediction

Systematic and transparent validation of a routine genetic evaluation in pigs using MiXValidate

---

Abstract

Even properly designed genetic evaluations need to be validated periodically. Validation studies, however, are notoriously laborious, cumbersome, prone to mistakes, and open to diverse interpretation, especially for indirect genetic effects. MiXValidate was developed to facilitate carrying out and interpreting routine forward-in-time validation of genetic evaluations, involving direct and maternal genetic effects. From the full phenotypic data, pedigree and genotypes, MiXValidate creates a second dataset without phenotypes of validation individuals, their siblings, their progeny, and any descendants of these individuals. Pedigree and genotypes are the same. MiXValidate then automatically evaluates each dataset and uses solutions and adjusted phenotypes of validation animals to calculate validation statistics. They include those of the Linear Regression (LR) method for cross-validation and adjusted-phenotype validation. It is possible to either validate parent genetic effect solutions with progeny performance, or individual genetic effect solutions with the performance of the individual itself. Validation of indirect genetic effects, such as maternal genetic effects, are also supported. They require a separate evaluation because data selection differs from the validation of direct genetic effects. We demonstrate MiXValidate features and its application using a dataset of performance test traits of purebred pigs of a commercial boar line. The full dataset (FULL) included animals born from October 2014 onwards (10 years), with a total of 227,107 animals with a phenotype for at least one trait. The total number of genotypes was 162,790. All genotyped animals also have phenotypes. Validation animals were genotyped animals born between October 2024 - October 2025 (N = 11,524). The partial data (PARTIAL) consisted of the full dataset with phenotypes of validation animals and their siblings removed. None of the validation animals or their siblings had progeny. Performance test traits included in this study are moderately heritable (~0.40). Validation statistics for a trait were calculated using only data of validation animals with a valid phenotype for that trait. The slope of solutions of FULL on solutions of PARTIAL is a measure of dispersion bias and was close to the expectation of 1.00. Intercept and level bias were less informative because solutions of FULL and PARTIAL were not corrected to a fixed genetic base. Realised prediction accuracies were high and ranged from 0.674 to 0.847. This MiXValidate evaluation only required a list of suitable validation individuals to validate the routine genetic evaluation. Even in multi-breed evaluations, validation individuals should be chosen from a single breed or cross to avoid inflated realised prediction accuracies. Because descendants of validation animals and descendants of their siblings are also excluded from the partial data, MiXValidate allows to do validation studies retrospectively. MiXValidate offers the opportunity of a user-friendly validation process for large routine (genomic) evaluations, facilitating a consistent and transparent validation.

Keywords: 2026

How to Cite:

Bonifazi, R., Schrauf, M., ten Napel, J., van der Spek, D. & Vandenplas, J., (2026) “Systematic and transparent validation of a routine genetic evaluation in pigs using MiXValidate”, World Congress on Genetics Applied to Livestock Production Digital Archive 2026(1): 2284536. doi: https://doi.org/10.31274/wcgalp.23580

Rights: 1