|
Abstract:
|
Most genome-wide association studies (GWASs), including Alzheimer's Disease Neuroimaging Initiative (ADNI), are based on the case-control study design. Without adjusting for the bias within sample design and fitting the regression models on it directly to detect the association between some certain genetic factors and secondary phenotypes may result in some invalid consequence such as biased coefficient estimations or inflated Type 1 error. Many methods are proposed in previous study to correct the biased when the group status (primary trait) is binary. However, some data samples may have multiple primary traits like ADNI is grouped by : Alzheimer's Disease(AD). mild cognitive impairment (MCI) and normal controls. So the aim of this article is to extend the bias correction from simple binary case-control sample to multiple-group samples by generating a more general framework. The idea behind the framework is to fit a transformed model which is approximately random based on the biased sample when the estimated proportions of each group in the real world is known. GWAS with ADNI data using the framework detects more genetic effects than the regression model without bias correction.
|
