PedImpute

PedImpute performs haplotype construction and genotype imputation for whole-genome single nucleotide polymorphism (SNP) panels in complex pedigrees to support genomic evaluations in dairy cattle.

Key Features:

• Implementation: Fortran-based program implementation for computational processing of genotype data.
• Target data and pedigrees: Operates on whole-genome SNP panels and is suited to complex pedigrees including medium to large half-sib families typical of dairy cattle.
• Algorithm efficiency: Balances imputation accuracy and computational efficiency to enable routine genomic evaluations with reduced computing time.
• Comparative performance: Demonstrated performance on 19,904 Holstein genotypes from an international dataset (United States, Canada, UK, Italy) and showed superior results when sire, dam, and maternal grandsire were genotyped at high density compared with findhap and BEAGLE.
• Accuracy metrics: Reported error rates between 0.2% and 1.1% and allelic squared correlation values between 96.6% and 99.3%.
• Scenario adaptability: Excels when key ancestors (sire, dam, maternal grandsire) are genotyped at high density, while BEAGLE can outperform PedImpute in scenarios with low-density or absent genotypes among close relatives but with higher computational demands.

Scientific Applications:

• Genomic evaluations: Imputation and haplotype construction to support genomic evaluations in dairy cattle breeding programs.
• Genetic selection and breeding decisions: Providing imputed genotypes for use in selection decisions where complete ancestral genotyping improves prediction accuracy.
• Large-scale dataset handling: Application to large genotype datasets such as multi-country Holstein collections to improve accuracy of genomic predictions.

Methodology:

Fortran implementation of a novel algorithm that constructs haplotypes and imputes genotypes by leveraging high-density genotyping data from pivotal ancestors (sire, dam, maternal grandsire) to optimize speed and precision.