Visualizing Relationships between Conformation Traits over Time

Abstract

Genetic evaluations often include many correlated traits, which are necessary in selection indices. The genetic parameters underlying their estimation have the potential to change at an accelerated pace under genomic selection. Standard methods for their evaluation are computationally more expensive with large numbers of animals over several years. This study aims to present a methodology to visualize relationships between conformation traits over time. Genetic Parameters via Predictivity is a method that allows this study to graph correlations among 18 linear conformation traits over yearly times slices from 2010 through 2020 using all available genomic, pedigree and phenotypic information For the initial time slice of 2010, the validation population included cows born in 2009 through 2011 while the reference population included cows born from 2000 through 2008. Each subsequent time slice increases the year by one. Using the most recent time slice, with a validation subset of cows born between 2019 through 2021, each trait was regressed on every combination of 2, 3, and 4 remaining traits, choosing the combination with the highest explained variance (R²). The chosen combinations were then used to compute and graph the predictor traits for all time slice going back to 2010. No more than 4 predictor traits were needed for each target trait, as evidenced by the low 4^th eigenvalue in 2010 and 2020, where principal component analysis was used to graph biplots.