Improved Allele Frequencies in gnomAD through Local Ancestry Inference

The Genome Aggregation Database (gnomAD) has been an invaluable resource for the genomics community, quantifying allele frequencies across multiple genetic ancestry groups. However, traditional gnomAD allele frequencies report aggregate estimates, overlooking fine-scale genetic differences. This can skew allele frequency estimates, particularly in recently admixed populations, and potentially lead to clinically relevant misinterpretations in underrepresented groups. To address this, we have conducted local ancestry inference (LAI) in two gnomAD genetic ancestry groups with high degrees of recent admixture, Admixed American (n=7,612) and African/African American (n=20,250),to derive more precise genetic ancestry-specific allele frequencies, enhancing the granularity and accuracy of frequency estimates. We demonstrate that LAI enables identification of important disease-risk variants that occur at higher frequencies within specific genetic ancestry components. Additionally, we found that 85.1% of variants in the African/African American and 78.5% in the Admixed American genetic ancestry groups show at least a twofold difference in ancestry-specific allele frequencies and 81.49% of sites analyzed would receive an increased grpmax value. With the release of this new LAI data, we provide a higher-resolution view of allele frequency across specific ancestry backgrounds, which is obscured when assessing frequency at the aggregate group level, enabling more precise and accurate genomic interpretations.