Rare Variants Imputation in Admixed Populations: Comparison Across Reference Panels and Bioinformatics Tools

Area of Science:

• Genomics
• Population Genetics
• Bioinformatics

Background:

• Genome-Wide Association Studies (GWAS) commonly use imputation for untyped markers.
• Imputation accuracy for common variants is established, but less is known for rare and ultra-rare variants.

Purpose of the Study:

• To assess the imputation accuracy of rare and ultra-rare variants in an admixed Caribbean Hispanic population (CH).
• To compare the performance of different reference panels, phasing, and imputation algorithms.

Main Methods:

• Evaluated imputation accuracy in 1,000 CH individuals for rare (0.1%-1% MAF) and ultra-rare (<0.1% MAF) variants.
• Utilized Haplotype Reference Consortium (HRC) and 1000 Genomes Project (1000G) reference panels with SHAPEIT/Eagle2 phasing and IMPUTE2/MACH-Admix imputation algorithms.
• Assessed quality using internal indexes, Wilcoxon Signed-Rank Test, Cohen's kappa coefficient, and imputation of a specific mutation (PSEN1 G206A) with confirmation genotyping.

Main Results:

• SHAPEIT-IMPUTE2 with the HRC panel yielded higher imputation quality for rare and ultra-rare variants compared to 1000G.
• SHAPEIT2 outperformed Eagle2 in retrieving high-quality imputed variants.
• African ancestry was significantly associated with increased imputation errors for rare variants (p < 1e-05).

Conclusions:

• Larger haplotype reference panels enhance imputation accuracy for rare and ultra-rare variants in admixed populations.
• Population's ethnic composition, specifically African ancestry, is a key predictor of imputation accuracy, with higher proportions linked to poorer performance.