Multi-omics analysis

Estimating many variance components simultaneously using factor analytical models; using the rumen microbiome to predict the breeding value for methane

---

Abstract

Microbial features of the rumen microbiome have been proposed as indicators to improve breeding value prediction of methane emissions in dairy cattle. However, incorporating high-dimensional microbiome data into genetic evaluations remains a challenge because standard quantitative methods require estimation of many (co)variance components. The combination of a small dataset with many (microbial) traits is expected to lead to non-positive definite matrices (Hill and Thompson, 1978. Biometrics 34:429-439), and therefore poor convergence when estimating variance components. This study aimed to evaluate the use of factor analytic (FA) models as a parsimonious alternative to estimate genetic parameters between methane and multiple microbial features simultaneously. Rumen microbial data was available from 902 genotyped Holstein cows of which 645 had both methane records and microbiome samples. Thirty microbial taxa were selected a priori based on their heritability and genetic correlation with methane records. Variance components were estimated using ASReml, fitting hierarchical FA models of increasing order (XFA(k), k = 1- n), where the genetic covariance matrix S=LL′+P, where L is a matrix of k loadings on the covariance scale and P is a diagonal vector of specific variances. Model fit was assessed using Akaike's Information Criterion (AIC). The full genetic matrix was used to predict accuracy of breeding values for methane using different combinations of microbial features, using selection index theory. Model fit improved with increasing FA order, with the optimal model identified as XFA7, that has 248 (co)variances explaining the full matrix with 465 genetic (co)variances. Heritability of methane was moderate, and heritabilities for microbial taxa ranged from 0.16 to 0.77. Genetic correlation among microbial taxa were on average 0.17 (sd = 0.49), and with methane ranging between -0.63 and 0.38, with an average of -0.21 (sd = 0.26). The combined use of microbial features to improve the predicted accuracy of methane breeding values compared with the accuracy of using methane records (i.e. the square root of the heritability of methane) alone will be assessed using selection index methodology. Factor analytic models provide a computationally efficient and statistically robust framework for simultaneously estimating large genetic covariance matrices involving microbiome traits and methane. Treating microbial features as quantitative phenotypes allows integration into existing genetic evaluation systems, and quantitative genetics theory to evaluate the impact on selection for reduced methane emissions within breeding programs.

Keywords: 2026

How to Cite:

Veerkamp, R., Honerlagen, H., Flossdorf, D., van Breukelen, A., Manzanilla-Pech, C. & Bouwman, A., (2026) “Estimating many variance components simultaneously using factor analytical models; using the rumen microbiome to predict the breeding value for methane”, World Congress on Genetics Applied to Livestock Production Digital Archive 2026(1): 2286299. doi: https://doi.org/10.31274/wcgalp.23912

Rights: 1