# Predicting productive, health, and reproductive traits in Mexican Holstein cattle using single nucleotide polymorphisms, haplotypes, and runs of homozygosity

## TL;DR

This study explores how genomic data can improve predictions of productivity, health, and reproduction in Mexican Holstein cattle.

## Contribution

The study compares the effectiveness of SNPs, haplotypes, and ROH in predicting traits and finds that ROH provides higher genetic variance estimates.

## Key findings

- Using ROH increases the estimation of genetic variance for health and reproductive traits.
- SNPs provided the highest predictive correlation for traits like milk yield and somatic cell score.
- Combining SNPs, HAP, and ROH did not improve predictive performance over using SNPs alone.

## Abstract

Summary: We investigated the use of different sources of genomic information, such as single nucleotide polymorphisms (SNPs), haplotypes (HAP), and runs of homozygosity (ROH), to predict productive, health, and reproductive traits in Mexican Holstein cattle. The use of models including genomic information from DNA markers increased the predictive performance and yielded higher estimates of genetic variance compared with models that included only fixed effects of complex traits in Mexican Holstein cattle.

Summary: We investigated the use of different sources of genomic information, such as single nucleotide polymorphisms (SNPs), haplotypes (HAP), and runs of homozygosity (ROH), to predict productive, health, and reproductive traits in Mexican Holstein cattle. The use of models including genomic information from DNA markers increased the predictive performance and yielded higher estimates of genetic variance compared with models that included only fixed effects of complex traits in Mexican Holstein cattle.

•The use of ROH increases the estimation of genetic variance for health and reproductive traits.•The combination of SNP, HAP, and ROH does not increase predictive performance.•Kernel-based models (Bayesian) can capture greater genetic variance for complex traits.

The use of ROH increases the estimation of genetic variance for health and reproductive traits.

The combination of SNP, HAP, and ROH does not increase predictive performance.

Kernel-based models (Bayesian) can capture greater genetic variance for complex traits.

The aim of this study was to evaluate the use of 3 different genomic relationship matrices built from SNPs, haplotypes (HAP), and runs of homozygosity (ROH), on phenotype predictive ability and estimated genetic variance of milk yield, SCS, and days open in Mexican Holstein cattle. The analyses included the use of the genomic relationship matrices as kernel-based models fitting either one or multiple sources of information. The SNPs and HAP matrices were built as linear kernels, and the ROH matrix as a Gaussian kernel. Also, we used as a reference the performance of the single-step GBLUP. Predictive ability was evaluated in 10-fold cross-validation. The highest predictive correlation was obtained using SNPs (0.63 for SCS, 0.57 for milk yield, and 0.20 for days open). The use of multigenomic relationships, including HAP and ROH, did not outperform the use of only SNPs in predictive ability, but the highest genetic variance was estimated using ROH (0.39 for milk yield, 0.26 for SCS, and 0.22 for days open).