Analysis of CFTR Variants in the All of Us Genomic Database by Genetic Ancestry
Lilly Bar
Mentor: Professor Markus Hoffmann, Georgetown University Medical Center.
Date/Time: August 25th, 2026 at 3:15 PM.
Abstract: Background: Cystic fibrosis (CF), an autosomal recessive genetic disorder, results from pathogenic variants in the cystic fibrosis transmembrane conductance regulator (CFTR) gene. CF can be life-threatening if not diagnosed and treated, but advancements in newborn screening and therapies are promising, especially for addressing pathogenic variants in persons of non-European ancestry. The analysis of large genomic databases can accelerate the discovery of pathogenic variants in persons with rare diseases, and thus inform strategies in disease diagnosis and treatment. It is crucial to evaluate the ability of computational tools to identify and analyze variants associated with genetic diseases, especially as artificial intelligence becomes increasingly integrated in the life sciences.
Methods: CFTR variant frequencies were obtained from the NIH’s All of Us Research Program database (v9). CFTR variants were filtered by genetic ancestry groups: African (AFR), Admixed American (AMR), East Asian (EAS), European (EUR), Middle Eastern (MID), South Asian (SAS), and Other (OTH). CFTR variants were further filtered by pathogenic and likely pathogenic ClinVar classifications. CFTR variants with allele counts greater than 20 were analyzed by descending frequency. Variants and their corresponding frequencies were then compared to CFTR variants captured in the Genome Aggregation Database (gnomAD, v4) per genetic ancestry group. ACMG’s expanded list of variants for CFTR carrier screening was used to evaluate current carrier screening. The responsiveness of pathogenic or likely pathogenic variants to Alyftrek was also evaluated. AlphaMissense pathogenicity scores were used to evaluate the ability of AlphaMissense to predict pathogenicity in the context of a rare genetic disease.
Results: Frequencies of filtered pathogenic and likely pathogenic CFTR variants ranged from 1.16e-4 to 2.78e-3 (AFR), 1.03e-4 to 4.73e-3 (AMR), 3.5e-5 to 1.43e-2 (EUR), 2.4e-3 to 3.02e-3 (SAS), and 2.82e-4 to 7.661e-3 (OTH). By genetic ancestry group, 10.7 (AFR), 12.6 (AMR), 2.9 (EUR), 7.1 (OTH), and 0% (SAS) of filtered alleles were absent from ACMG’s carrier panel. 12.8-26.8% of filtered variants were unresponsive to Alyftrek among participants with African, Admixed American, European, and other genetic ancestry. AlphaMissense pathogenicity scores ranged, despite corresponding to pathogenic variants appearing at relatively high frequencies across genetic ancestry groups.
Conclusion: Pathogenic and likely pathogenic CFTR variants listed in the All of Us database are similar to those in gnomAD, and further, appear at comparable frequencies. Though ACMG’s set of recommendations improves carrier screening immensely, the detection of filtered alleles was less effective for persons of African and Admixed American genetic ancestry than European ancestry. An analysis of CFTR variants with allele counts fewer than 20 would further aid in evaluating the performance of current panels and therapies.