Genotype imputation performance in Nellore cattle across different SNP panels and software tools

TL;DR

This study shows that genotype imputation works well in Nellore cattle using commercial SNP panels and various software tools, with FImpute performing best.

Contribution

The study evaluates imputation performance in Nellore cattle using multiple software tools and SNP panels, a less-studied population.

Findings

FImpute showed the highest accuracy and consistency across different SNP panel densities.

Imputation accuracy improved with higher panel density but remained challenging for rare alleles (MAF ≤ 0.03).

All tested software achieved high imputation accuracy, making genotype imputation feasible for Nellore cattle.

Abstract

Most commercial genotyping panels were designed for Bos taurus taurus breeds, which may reduce their informativeness in Bos taurus indicus populations like Nellore due to differences in linkage disequilibrium and minor allele frequencies. Genotype imputation is a cost-effective method to increase marker density by predicting missing genotypes in low- and medium-density (LD and MD) SNP panels using high-density (HD) reference data. While extensively studied in Bos taurus taurus breeds, limited data exist for Nellore cattle. This study evaluated the imputation performance of four widely used software tools—FImpute, Beagle, Minimac, and Findhap—using a large dataset of over 305,000 Nellore animals genotyped by the Brazilian Association of Zebu Breeders (ABCZ). Imputation was performed in two steps: from LD/MD panels to a customized 120k panel, then to a 777k HD panel. Accuracy was measured using Pearson correlation (Corr), the percentage of correctly imputed genotypes (PERC), and computational performance. All software achieved high accuracy (Corr: 0.82–0.98; PERC: 89.35%–99.26%). FImpute consistently showed the best performance, with the highest Corr (0.983), the fewest poorly imputed individuals, and strong results across all panel densities. While pedigree information in FImpute offered some benefits, its overall improvement compared to population-based methods was limited in this study. Imputation accuracy improved with panel density and was stable across most MAF ranges, though rare allele (MAF ≤ 0.03) imputation remained challenging. These findings confirm that accurate genotype imputation in Nellore cattle is feasible using existing commercial SNP panels.

The online version contains supplementary material available at 10.1007/s11250-026-04914-0.

Figures