Premium
Polygenic risk score for Alzheimer’s disease in Caribbean Hispanics
Author(s) -
Sariya Sanjeev,
Felsky Daniel,
Lali Ricky,
ReyesDumeyer Dolly,
Lantigua Rafael A,
Brickman Adam M.,
Manly Jennifer J,
Vardarajan Badri N.,
JiménezVelázquez Ivonne Z.,
Pare Guillaume,
Mayeux Richard,
Tosto Giuseppe
Publication year - 2021
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1002/alz.055031
Subject(s) - cohort , logistic regression , medicine , demography , receiver operating characteristic , alzheimer's disease , genetic association , disease , apolipoprotein e , gerontology , biology , single nucleotide polymorphism , genetics , genotype , gene , sociology
Abstract Background Polygenic risk scores (PRS) provide an overall estimate of the individual’s genetic propensity to a trait by combining sparse information scattered across multiple genetic loci which often display small effect sizes. Most genetic studies are of European ancestry, ultimately limiting the use of PRS in other ethnicities. Here we constructed and validated a PRS for late‐onset Alzheimer’s Disease (LOAD) in Caribbean Hispanics (CH). Method We employed genome‐wide association summary statistics from 4,312 CH to construct an ancestry‐specific PRS (“CH‐PRS”) in an independent validation CH cohort (N=1,850). CH‐PRS performance was evaluated using the area under the receiver operator characteristic (ROC) curves and logistic regression to evaluate strength of the association with LOAD and statistical significance. We sought to further replicate the CH‐PRS in an independent CH dataset (Alzheimer’s Disease Research Center “ADC‐CH”, N=200) and in a brain autopsy cohort (N=33). We also studied the CH‐PRS performance in predicting conversion to LOAD in a subset of non‐demented individuals at baseline with longitudinal data (N=600), employing a Cox regression model. Finally, we tested the effect of ethnicity on PRS performances by employing European (EUR) and African American (AA) ancestry studies as discovery datasets to construct alternative PRSs (“EUR‐PRS”, “AA‐PRS”) in our validation cohort. Result In the full model (LOAD ∼ CH‐PRS + sex + age + APOE‐ɛ4) , the AUC reached 74.02% (OR=1.51 95%CI=1.36‐1.68), raising to >75% in APOE‐ɛ4 non‐carriers. In the autopsy cohort, higher CH‐PRS was significantly associated with pathological AD diagnosis (AD ∼ CH‐PRS; AUC=72%; OR=2.35, 95%CI=1.0‐5.52), and AUC=83% in the full model. In ADC‐CH, the PRS showed significant association with LOAD (OR=1.61, 95%CI=1.19‐2.17). CH‐PRS significantly predicted conversion to LOAD over time (HR=1.96, CI=1.61‐2.39). EUR‐PRS and AA‐PRS reached lower prediction accuracy (AUC=58% and 53%, respectively). Conclusion PRS is an effective strategy to delineate individual risk profiles as shown by our cross‐sectional and longitudinal analyses as well as associations with well‐established AD‐hallmarks. Enriching diversity in genetic studies is indeed critical to provide a PRS tool that is effective in predicting LOAD risk across ethnic populations.