Basic Science and Pathogenesis.

Area of Science:

• Genomics
• Population Genetics
• Bioinformatics

Background:

• Genomic studies rely on imputation to infer untyped variants, crucial for genome-wide association studies (GWAS).
• Imputation accuracy can be variable, particularly for rare variants and across different ancestral populations.
• The Alzheimer's Disease Sequencing Project (ADSP) provided genotype data for this study.

Purpose of the Study:

• To compare the accuracy of traditional single-round imputation (SI) versus double imputation (DI) using whole genome sequencing (WGS) data.
• To evaluate imputation error rates across various minor allele frequency (MAF) categories and ancestral groups.

Main Methods:

• Utilized ADSP genotype data from 196 Caribbean Hispanics.
• Estimated imputation error rates for single nucleotide polymorphisms (SNPs) with high quality (R^2≥ 80%) on chromosome 1.
• Analyzed error rates across MAF brackets (common, uncommon, rare, ultra-rare) and for individuals with high African ancestry (AFR).

Main Results:

• Double imputation (DI) demonstrated significantly lower overall error rates (2.65%) compared to single imputation (SI) (4.23%).
• DI consistently outperformed SI across all MAF categories, including rare and ultra-rare variants.
• Individuals with higher African ancestry showed increased imputation errors, but DI still provided superior accuracy in this group.

Conclusions:

• Multiple rounds of imputation (DI) offer enhanced accuracy over traditional single-round imputation (SI).
• This improvement is particularly notable for rare and ultra-rare variants and in populations with higher imputation error rates, such as those with African ancestry.
• DI represents a valuable and accessible method for improving imputed data quality, potentially leading to more robust genetic association studies.