The Impact of Model of Inheritance to Different Family-based Haplotype Association Tests― Using QTL trait as the Outcome

• Main Authors: Yu-Chun Kuo, 郭育君
• Other Authors: Cathy SJ Fann PhD
• Format: Others
• Language: en_US
• Published: 2006
• Online Access: http://ndltd.ncl.edu.tw/handle/92606790410253397554

碩士 === 國立陽明大學 === 遺傳學研究所 === 94 === Complex diseases are caused by multiple-gene (multiple genes) and their interaction with environmental factors, and they account for the relatively high percentage of overall human diseases. In order to understand the pathogenesis of complex diseases, researchers are devoted to identify disease causing genes, therefore more detailed biological mechanisms can be understood for future genetic consultation. Due to the rapid increases of available genetic markers and statistical software, family-based haplotype association analysis has now become one of the common strategies to identify disease genes. Complex diseases belong to the non-Mendelian inheritance pattern, and many quantitative characters can contribute to the pathogenesis. For example, hypertriglyceridaemia and high concentration of low density lipoprotein cholesterol (LDL-C) in the blood are involved in various mechanisms of hypertension development. Therefore, using quantitative trait can obtain more genetic information as compared with using dichotomous trait (affected/unaffected). Up to our knowledge, QPDTPHASE and HaploFBAT are the two available programs to perform haplotype-based quantitative trait association tests. These two algorithms use different strategies to handle the missing parental haplotype phase data, therefore inconsistent results are often observed for those late onset diseases where parental data are usually missing. The objective of our investigation is to compare the power and type I error for these two programs to clarify their advantages and disadvantages in quantitative trait analyses. Simulations were considered for the following parameters: 1) various parental data missing rates, 2) various proportions of two subpopulations, 3) different genetic inheritance models, 4) variance of the trait in the subpopulations and 5) the genetic effect imposed on the trait. The results indicate that population stratification is the major factor to cause discordance between these two algorithms. If subpopulation ratio is 1:1, HaploFBAT is always more powerful than QPDTPHASE, but type I error rate of HaploFBAT program is 0.071~0.097(α=0.05) when missing rate of parental genotypes reaches 80%. If degree of population stratification is unknown, QPDTPHASE may be the better choice. The type I error estimated by these programs is close to the pre-specified level in most datasets. For future usage, our findings can be used to explain for the discrepant results between these programs and provide a theoretical foundation for researchers in choosing appropriate program to study the QTL traits of interest.