請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/27002
標題: | 遺傳相關研究中族群分層偏差之剖析 Anatomy of the Population Stratification Biases in Genetic Association Studies |
作者: | Liang-Yi Wang 王亮懿 |
指導教授: | 李文宗(Wen-Chung, Lee) |
關鍵字: | 族群分層偏差,病例對照研究法,唯病例法,長壽基因研究,橫斷式研究法,暴露配對,族層配對, population stratification bias,case-control study,case-only study,longevity study,cross-sectional study,exposure matching,stratum matching., |
出版年 : | 2008 |
學位: | 博士 |
摘要: | 相關分析是近年來易感受性基因研究常用的方法。利用相關分析探討基因對疾病的效應時,若收集一般族群做為對照,可能會產生潛在的族群分層偏差,造成研究結果的誤判。因此本論文利用公式的推導,針對過去未深入探討的議題,進行基因相關研究的族群分層偏差剖析:(1)病例對照研究法中,族群分層偏差對基因主效應的影響;(2)唯病例法中,族群分層偏差對基因環境交互作用的影響;(3)長壽基因橫斷式研究中,族群分層偏差對基因主效應的影響;(4)暴露配對與族層配對對族群分層偏差控制的效果。本論文有趣的發現,上述族群分層偏差皆由相似的 模組所堆砌而成。利用唯病例法估算基因環境交互作用時,在合理的情境下就常見5%以上的族群分層偏差,影響程度高於病例對照研究中族群分層偏差對基因主效應的影響。本論文並提供簡單的公式,讓研究者計算研究可能產生的族群分層偏差極值,以利研究結果的解釋。我們也發現長壽基因橫斷式研究,除了過去此類研究所擔心的世代效應偏差,另有因為分層族群長期追蹤,以及因為分層族群跨代人口數變動而產生的族群分層偏差。我們因此認為橫斷式研究所產生的偏差會很嚴重。過去基因研究者為聚焦於基因效應的探討,常配對環境暴露因子(以求去除不相干的暴露效應)。我們發現此等暴露配對有時會造成偏差量不降反升。由於族層配對不會有此等反效果,因此我們提倡配對族層而提醒研究者不要不假思索逕行配對暴露。 The genetic association study is a popular method for evaluation of susceptibility genes. Using population-based controls may suffer from bias due to population stratification. Population stratification bias may cause over- or underestimation of a genotype relative risk. In this paper, we address the following issues: (1) population stratification bias in a case-control study for genetic main effects, (2) population stratification bias in a case-only study for gene-environment interactions, (3) population stratification bias in a longevity cross-sectional study for genetic main effects, and (4) the matching effectiveness of exposure-matching vs. stratum-matching. It is of interest to find that the population stratification biases in various situations are all composed of similar module of the form . The estimated interaction effect in a case-only study of realistic scenarios is frequently biased by more than 5%, which is larger than the bias of genetic main effect in a case-control study. Researchers can use the boundary formulas for population stratification bias provided in this paper to make more prudent interpretations of their results. It is found that the cross-sectional design in a longevity study, besides the “cohort bias” previously described, can suffer from bias due to long-term follow-up in a stratified population and bias due to changes in population numbers between cohorts in a stratified population. Therefore a cross-sectional longevity study is prone to be seriously biased. Matching on exposures is a common practice when interests are focused on the effects of genes (in the hope that such doing will remove the irrelevant exposure effects). We found that exposure matching can sometimes increase the magnitude of bias instead of reducing it. By contrast, matching on stratum will guarantee the magnitude of bias to be smaller than that of an unmatched study. Thus we promote stratum matching and caution against uncritical use of exposure-matching in a genetic association study. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/27002 |
全文授權: | 有償授權 |
顯示於系所單位: | 流行病學與預防醫學研究所 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-97-1.pdf 目前未授權公開取用 | 537.51 kB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。