fastPHASE

fastPHASE estimates missing genotypes and haplotypes in genetic datasets by clustering haplotypes to model local linkage disequilibrium for genotype imputation and haplotype phasing.

Key Features:

• Clustered haplotype model: Clusters haplotypes into groups of similar sequences over short chromosomal regions to reflect the local nature of recombination.
• Hidden Markov model: Uses a hidden Markov model that allows cluster memberships to change continuously along the chromosome.
• Genotype imputation: Performs genotype imputation with accuracy reported as comparable to or surpassing other methodologies.
• Haplotype estimation (phasing): Produces haplotype phase estimates with a reported switch error rate of 0.055 versus 0.051 for PHASE on unrelated HapMap individuals.
• Uncertainty calibration: Computes well-calibrated probabilities that reflect uncertainty in genotype and haplotype estimates.
• Scalability and efficiency: Designed for large-scale datasets, e.g., thousands of individuals genotyped at hundreds of thousands of markers, with substantially reduced computational demands relative to some methods.

Scientific Applications:

• Genotype imputation: Imputes missing SNP genotypes in large-scale datasets to support association studies and downstream analyses.
• Haplotype phasing: Estimates haplotypic phase in unrelated individuals for linkage disequilibrium and haplotype-based analyses.
• Population genetics: Characterizes local LD patterns and recombination structure in population-genetic studies.

Methodology:

Clusters of haplotypes are modeled with a hidden Markov model that permits cluster membership changes along the chromosome, accommodating block-like LD patterns and gradual declines in LD, and computes calibrated probabilities for estimates.