Haplotypimputation

Haplotypimputation, or haplotype imputation, is the statistical inference of unobserved genetic variation, particularly haplotypes or missing genotypes, in a set of samples using observed genotype data and a reference panel of haplotypes. It exploits linkage disequilibrium, the non-random association of alleles along the chromosome, to predict which alleles co-occur on the same chromosome.

Most methods first phase individuals to infer haplotype structure, then impute missing genotypes by comparing the individual's haplotypes to a reference panel. The approach is commonly implemented with hidden Markov models that treat an individual's genome as a mosaic of reference haplotypes. Popular software includes BEAGLE, IMPUTE2/IMPUTE5, Minimac4, and SHAPEIT for phasing. Reference panels such as the 1000 Genomes Project, the Haplotype Reference Consortium (HRC), and TOPMed provide dense haplotype catalogs.

The output includes genotype dosages or probabilistic calls, which can be used directly in downstream analyses or converted to hard genotypes. Imputation increases variant density, harmonizes data across studies, and boosts power in genome-wide association studies and fine-mapping, as well as in polygenic risk scoring.

Accuracy depends on the representativeness of the reference panel, the ancestry of the study sample, and allele frequency; imputation is less reliable for rare variants or in regions with complex LD. Quality metrics such as Rsq or info score gauge reliability. Computational cost and data privacy considerations also apply.

Haplotypic imputation is standard in modern human genetics and is applicable to plant and animal genetics as well, enabling cost-effective genotyping and integrative analyses across datasets. Ongoing work aims to expand diverse reference panels and improve phasing and imputation in admixed populations.