Imputation accuracy to whole-genome sequence in Nellore cattle.

Area of Science:

• Animal Genetics
• Genomic Selection
• Bioinformatics

Background:

• Cost-effective DNA sequencing for animal genetic evaluation relies on imputing genotypes from SNP panels using key ancestor sequences.
• Imputation accuracy is crucial, especially for low-frequency causal variants within genes or regulatory regions.

Purpose of the Study:

• To assess whole-genome sequence imputation accuracy in Nellore beef cattle.
• To evaluate imputation performance for variants across different functional annotation classes.

Main Methods:

• Used genotype data from 151 sequenced Nellore sires and ~10,000 genotyped animals.
• Employed FImpute3 and Minimac4 (with Eagle phasing) for genotype imputation.
• Evaluated accuracy via fivefold cross-validation, measuring squared correlation between observed and imputed genotypes.

Main Results:

• Achieved high average per-animal imputation accuracies (0.94 for FImpute3, 0.95 for Minimac4).
• Minimac4 excelled with common variants (MAF > 0.03), while FImpute3 was better for low-frequency variants (MAF ≤ 0.03).
• Both methods showed high SNP-wise accuracy across all annotation classes; Minimac4 Rsq predicted accuracy well.

Conclusions:

• Whole-genome sequence imputation is feasible in Nellore cattle with high individual accuracy.
• SNP-wise imputation accuracy varies by software, particularly for rare variants.
• Imputation accuracy is largely independent of variant annotation class.