FImpute

FImpute imputes missing genotypes at untyped loci in large-scale livestock populations by leveraging high-density reference genotypes and pedigree or population haplotype sharing to support genomic selection.

Key Features:

• Deterministic Algorithm: Employs a deterministic algorithm instead of stochastic Hidden Markov Models (HMMs) to reduce computational demands while maintaining high accuracy.
• Overlapping Sliding Window: Uses overlapping sliding windows that start large to capture long-range haplotypes and progressively reduce in size to detect more distant relationships across chromosomes.
• Family and Population Integration: Integrates pedigree (family) information when available and broader population data to exploit shared haplotypes of varying length and frequency.
• Rare Variant Imputation: Demonstrates superior accuracy for imputing rare variants compared to some other methods.
• Comparative Performance and Scalability: Reports higher or comparable accuracy to Beagle and Impute2 in cattle datasets and can impute from 6 k to 50 k genotypes for 2,000 individuals in approximately 28 minutes using a reference panel of 64,429 individuals.

Scientific Applications:

• Genomic Selection: Imputes high-density genotypes to enable genomic selection and improve breeding program decision-making.
• Large-Scale Livestock Genotyping: Handles extensive datasets with hundreds of thousands of individuals genotyped on different panels for population-scale analyses.
• Rare Variant Studies: Supports the imputation and study of rare variants that are challenging to capture with other methods.

Methodology:

When available, the method begins with family-based imputation to establish initial haplotype matches; it then performs an overlapping sliding window search of varying sizes to identify and match long haplotypes in reference individuals, progressively reducing window size to capture more distant genetic relationships.