Proper selection of control set may minimize biases in case control association studies in Korean population.

Abstract

In a SNP association study on the human complex diseases, one of the most influential factors in finding significant markers is to select comparable case and control samples. In reality, however, there are many chances of selection biases in collecting the case and control samples. We have SNP association results for Korean stroke and schizophrenia patients vs. hospital controls. Since the controls in the study were healthy people who passed the health check-up in a local hospital, they might not represent general population. Thus, reliability of the association results should be reexamined using other unbiased control samples. To do this, we used the so-called KARE (Korean Association study REsources) data as an alternative control set, which were generated from 8,842 people in An-sung and An-san cohort in Korea. To determine the relative accuracy in detecting the significant markers, we adopted the public data related to the complex diseases as a sort of standard data set. We have found that the KARE set (cohort control) produces more replicable results than the original set (hospital control) in terms of consistency with the list of the already known diseases-causing markers.