Skip to content
Primary Menu

Education, Presentation, Publication

Funding & Recognition

Validating a Local Ancestry Inference Pipeline in Admixed Pacific Islander Cohorts 

Semester: Summer 2026


Presentation description

Pacific Islanders have some of the highest rates of obesity and type 2 diabetes yet are among the least represented in genomic research. Standard genome-wide association study (GWAS) methods correct for population structure via principal components or mixed models, but assume one averaged effect per variant. In admixed cohorts, this can dilute the power of variants restricted to one ancestral background, or those tagged through ancestry-specific linkage patterns. Local ancestry inference assigns each haplotype segment to an ancestral population, enabling ancestry-aware testing and recovering lost power. This pipeline has not previously been applied to a Pacific Islander cohort. We applied an RFMix-based local ancestry pipeline across three Pacific Islander cohorts (Utah Pacific Islander Diabetes Study (UPIDS), Samoan Adiposity Study (SAS), and Native Hawaiians in the MultiEthnic Cohort (MEC); n = 6,427) using a 300-individual reference panel (Oceanian, European, East Asian, African). We validated it via per-individual concordance with independent ADMIXTURE estimates, and recovery of a known population-specific BMI association with the CREBRF missense variant (rs373863828), comparing ancestry-aware TRACTOR to ancestry-naive analysis on identical cohorts and covariates. RFMix reproduced ADMIXTURE estimates in all three cohorts (Polynesian ancestry r = 0.983-0.992), from near-homogeneous SAS (96% Polynesian) to more admixed MEC (46-47%). For the CREBRF variant, both analyses returned similar effects (meta beta = 1.47 vs. 1.43), consistent with the published SAS estimate (1.36-1.45). Since the variant is near-absent outside Oceanian populations, non-Oceanian assignments are likely errors (15/2236; 0.67% error rate). The pipeline successfully recovers a known population-specific BMI-associated variant and produces ancestry calls matching independent estimates. This supports the possibility of discovering additional ancestry-specific or enriched cardiometabolic variants and extending this approach to biobank-scale admixed cohorts otherwise excluded from GWAS.
Presenter Name: William Vercillo
Presentation Type: Poster
Presentation Format: In Person
Presentation #26
College: Medicine
School / Department: Internal Medicine
Research Mentor: Marcus Pezzolesi
Additional Mentor: Travis Tu'ifua, Connor Littlefield
Time: 10:40 AM
Physical Location or Zoom link:

Henriksen