Multi-omics analysis

Integration of microbiome information to improve the prediction of phenotypes and breeding values in pig breeding

---

Abstract

Traditional prediction pipelines in animal breeding focus on genomic information as the primary type of omics data to include in prediction models. However, with recent technological advancements, other omics have become more accessible, and from a biological perspective, they should play a major role in important breeding traits. As inheritance patterns are less straightforward than those for genomic information, data generation is more costly, and datasets can highly depend on the time point or tissue type of collection. Practical use in breeding, therefore, comes with various additional challenges.In this study, we focus on the use of gut microbiome data collected from a panel of 1,412 pigs from two farms, which have been phenotyped for average daily gain, feed intake, backfat, and loin depth. For the prediction of phenotypes, the average prediction accuracy across the considered traits increased by 11% when using a linear mixed model that included separate random effects for the genomic and microbiome components, compared to a model that only included a genomic component. As genomic and microbiome data partially explain each other, and both components showed strong positive correlations for most traits, part of the genomic effect was accounted for in the microbiome component of the joint prediction model. To estimate a breeding value, the Genomic-Omics BLUP was used to predict the microbiome component based on genetics, effectively utilizing the microbiome data as an additional data source to reduce residual variance in the prediction model, resulting in an average increase in prediction accuracy of 5%.To mimic a real-world scenario where it is not possible to collect microbiome data from all individuals, an extended dataset of 6.691 animals was considered by imputing microbiome data through the use of genomic data in a single-trait linear mixed model by using the microbiome data of an individual operational taxonomic unit (OTU) as the trait to predict. Notably, the imputation approach improved the average prediction accuracy of phenotype predictions by 20% for backfat compared to the baseline when using only genomic data from the small panel. In comparison, the sole use of the additional phenotypes from the larger training population, without microbiota data, improved predictions by 11%, while adding microbiota data to the small training dataset increased accuracy by 16%. In the same setting, predictions for individuals with imputed microbiome data remained at a similar level, thereby providing a framework for utilizing partially available omics data.

Keywords: 2026

How to Cite:

Pook, T., Maruyama, T., Leite, N., Bergsma, R., Verschuren, L., Zemb, O. & Kamphuis, C., (2026) “Integration of microbiome information to improve the prediction of phenotypes and breeding values in pig breeding”, World Congress on Genetics Applied to Livestock Production Digital Archive 2026(1): 2284583. doi: https://doi.org/10.31274/wcgalp.23598

Rights: 1