Premium
A novel age‐informed approach for genetic association analysis in Alzheimer’s disease
Author(s) -
Guen Yann Le,
Belloy Michael E,
Napolioni Valerio,
Eger Sarah J,
Kennedy Gabriel,
Tao Ran,
He Zihuai,
Greicius Michael D
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.050541
Subject(s) - logistic regression , regression , proportional hazards model , regression analysis , disease , multivariate statistics , genetic association , exome , demography , medicine , statistics , biology , exome sequencing , genetics , genotype , single nucleotide polymorphism , mathematics , mutation , sociology , gene
Background The increased risk of Alzheimer’s Disease (AD) with age is well established. However, genome‐wide association studies of AD have often wrongly accounted for this known effect by covarying by age using age‐at‐onset for cases and age‐at‐last‐exam for controls. In most scenarios, this leads to controls being on average older than cases and the regression model incorrectly infers that age decreases AD risk. Methods Using simulated data, we compared the statistical power of several models: logistic regression on AD diagnosis adjusted and not adjusted for age; linear regression on a score integrating case‐control status and age; and multivariate Cox regression on age‐at‐onset. We applied these models to real exome‐wide data of 11,127 sequenced individuals (54% cases) and replicated suggestive associations in 21,631 genotype‐imputed individuals (51% cases) (Table 1). Results Modelling variable AD risk across age results in 10‐20% statistical power gain compared to logistic regression without age adjustment, while incorrect age adjustment leads to critical power loss (Figure 1). Applying our novel AD‐age score and/or Cox regression, we discovered and replicated novel variants associated with AD on KIF21B, USH2A, RAB10, RIN3 and TAOK2 genes (Figure 2, Tables 2, 3). Conclusion Incorrect age adjustment may remain unnoticed when the age difference between cases/controls is small and when the sample size is sufficient to identify causal variants with adequate statistical power. If the age difference is large, however, as in the whole‐exome sequencing of the ADSP for which controls are 10 years older than cases, then it is highly detrimental to adjust by age. Here we have proposed a novel AD‐age score to correctly integrate the age information in the phenotype and showed that it led to a gain in statistical power compared to a traditional logistic regression without age adjustment.