Fine-mapping in admixed populations using CARMA-X, with applications to Latin American studies
(2025) In American Journal of Human Genetics 112(5). p.1215-1232- Abstract
Genome-wide association studies (GWASs) in ancestrally diverse populations are rapidly expanding, opening up unique opportunities for novel gene discoveries and increased utility of genetic findings in non-European individuals. A popular technique to identify putative causal variants at GWAS loci is via statistical fine-mapping. Despite tremendous efforts, fine-mapping remains a very challenging task, even in the relatively simple scenario of studies with a single, homogeneous population. For studies with admixed individuals, such as within Latin America and the Caribbean, methods for gene discovery are still limited. Here, we propose a Bayesian model for fine-mapping in admixed populations, CARMA-X, that addresses some of the unique... (More)
Genome-wide association studies (GWASs) in ancestrally diverse populations are rapidly expanding, opening up unique opportunities for novel gene discoveries and increased utility of genetic findings in non-European individuals. A popular technique to identify putative causal variants at GWAS loci is via statistical fine-mapping. Despite tremendous efforts, fine-mapping remains a very challenging task, even in the relatively simple scenario of studies with a single, homogeneous population. For studies with admixed individuals, such as within Latin America and the Caribbean, methods for gene discovery are still limited. Here, we propose a Bayesian model for fine-mapping in admixed populations, CARMA-X, that addresses some of the unique challenges of admixed individuals. The proposed method includes an estimation method for the linkage disequilibrium (LD) matrix that accounts for small reference panels for admixed individuals, heterogeneity across populations and cross-ancestry LD, and a Bayesian hypothesis test that leads to robust fine-mapping when relying on external reference panels of modest size for LD estimation. Using simulations, we compare performance with recently proposed fine-mapping methods for multi-ancestry studies and show that the proposed model provides higher power while controlling false discoveries, especially when using an out-of-sample LD matrix. We further illustrate our approach through applications to two Latin American genetic studies, the Estudio Familiar de Influencia Genética en Alzheimer (EFIGA) study in the Dominican Republic and the Mexican Biobank, where we show the benefit of modeling ancestry-specific effects by prioritizing putative causal variants and genes, including several findings driven by ancestry-specific effects in the African and Native American ancestries.
(Less)
- author
- Yang, Zikun ; Wang, Chen ; Posadas-Garcia, Yuridia Selene ; Añorve-Garibay, Valeria ; Vardarajan, Badri ; Estrada, Andrés Moreno ; Sohail, Mashaal ; Mayeux, Richard and Ionita-Laza, Iuliana LU
- organization
- publishing date
- 2025-05-01
- type
- Contribution to journal
- publication status
- published
- subject
- keywords
- admixed fine-mapping, admixed GWAS, admixed populations, fine-mapping, GWAS, Latin American GWAS
- in
- American Journal of Human Genetics
- volume
- 112
- issue
- 5
- pages
- 18 pages
- publisher
- Cell Press
- external identifiers
-
- pmid:40147449
- scopus:105001147042
- ISSN
- 0002-9297
- DOI
- 10.1016/j.ajhg.2025.02.020
- language
- English
- LU publication?
- yes
- id
- 1e79369b-9562-400d-972c-2051fbf1e6e8
- date added to LUP
- 2025-09-10 11:18:17
- date last changed
- 2025-09-24 17:36:25
@article{1e79369b-9562-400d-972c-2051fbf1e6e8,
abstract = {{<p>Genome-wide association studies (GWASs) in ancestrally diverse populations are rapidly expanding, opening up unique opportunities for novel gene discoveries and increased utility of genetic findings in non-European individuals. A popular technique to identify putative causal variants at GWAS loci is via statistical fine-mapping. Despite tremendous efforts, fine-mapping remains a very challenging task, even in the relatively simple scenario of studies with a single, homogeneous population. For studies with admixed individuals, such as within Latin America and the Caribbean, methods for gene discovery are still limited. Here, we propose a Bayesian model for fine-mapping in admixed populations, CARMA-X, that addresses some of the unique challenges of admixed individuals. The proposed method includes an estimation method for the linkage disequilibrium (LD) matrix that accounts for small reference panels for admixed individuals, heterogeneity across populations and cross-ancestry LD, and a Bayesian hypothesis test that leads to robust fine-mapping when relying on external reference panels of modest size for LD estimation. Using simulations, we compare performance with recently proposed fine-mapping methods for multi-ancestry studies and show that the proposed model provides higher power while controlling false discoveries, especially when using an out-of-sample LD matrix. We further illustrate our approach through applications to two Latin American genetic studies, the Estudio Familiar de Influencia Genética en Alzheimer (EFIGA) study in the Dominican Republic and the Mexican Biobank, where we show the benefit of modeling ancestry-specific effects by prioritizing putative causal variants and genes, including several findings driven by ancestry-specific effects in the African and Native American ancestries.</p>}},
author = {{Yang, Zikun and Wang, Chen and Posadas-Garcia, Yuridia Selene and Añorve-Garibay, Valeria and Vardarajan, Badri and Estrada, Andrés Moreno and Sohail, Mashaal and Mayeux, Richard and Ionita-Laza, Iuliana}},
issn = {{0002-9297}},
keywords = {{admixed fine-mapping; admixed GWAS; admixed populations; fine-mapping; GWAS; Latin American GWAS}},
language = {{eng}},
month = {{05}},
number = {{5}},
pages = {{1215--1232}},
publisher = {{Cell Press}},
series = {{American Journal of Human Genetics}},
title = {{Fine-mapping in admixed populations using CARMA-X, with applications to Latin American studies}},
url = {{http://dx.doi.org/10.1016/j.ajhg.2025.02.020}},
doi = {{10.1016/j.ajhg.2025.02.020}},
volume = {{112}},
year = {{2025}},
}