Genotype imputation using the Positional Burrows Wheeler Transform.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Genotype imputation accuracy improves with larger reference panels.
• Large reference panels present significant computational challenges for imputation methods.
• Existing methods struggle to efficiently scale with increasing reference panel sizes.

Purpose of the Study:

• Introduce IMPUTE5, a novel genotype imputation method designed for scalability.
• Address the computational challenges posed by large-scale reference panels in genomics.
• Enhance the speed, accuracy, and memory efficiency of genotype imputation.

Main Methods:

• Utilizes the Positional Burrows Wheeler Transform (PBWT) data structure.
• Selects custom subsets of haplotypes for individual imputation.
• Identifies locally best matching haplotypes and identical-by-state segments.

Main Results:

• IMPUTE5 demonstrates significant speed improvements over MINIMAC4 (up to 30x) and BEAGLE5.1 (up to 3x).
• IMPUTE5 requires less memory compared to MINIMAC4 and BEAGLE5.1.
• The method exhibits sub-linear scaling with reference panel size, maintaining efficiency with millions of haplotypes.

Conclusions:

• IMPUTE5 offers a scalable and efficient solution for genotype imputation with large reference panels.
• The Positional Burrows Wheeler Transform is key to IMPUTE5's performance gains.
• IMPUTE5 facilitates accurate and computationally feasible genomic analyses using extensive haplotype data.