https://scholars.lib.ntu.edu.tw/handle/123456789/112616
標題: | 利用不確定性編碼矩陣進行多點遺傳相關研究 Marker-set Genetic Association Studies with An Uncertainty-coding Matrix |
作者: | 黃詠詳 Huang, Yung-Hsiang |
關鍵字: | 家族性研究;貝式統計;遺傳標記集合;單倍體;消失的遺傳性;family study;Bayesian;marker-set;haplotype;missing heritability | 公開日期: | 2012 | 摘要: | 相較於單點的遺傳標記分析研究,多點遺傳標記分析研究除了可能可以納入多點標記之間的生物關係,也較能提供有用的資訊且檢力較高的分析結果。例如,能考慮連鎖不平衡的單倍體就是一個多點遺傳標記的例子。只是,單倍體的分析仍然存在一些困難需要克服,比如說單倍體組成狀態的不明確、或者是在我們有興趣的某個基因區段的單倍體個數過多,甚至是在家族性研究中,親代傳遞單倍體給子代時的傳遞路徑也存在著不確定性。在本篇論文中,我將會分兩個階段來針對家族性研究提出一種不確定性編碼矩陣來解釋這些不確定性。首先,我會以三元體資料為主,以貝式羅吉斯迴歸模式來找出單倍體特異的估計值;其次,我會使用貝式廣義混合模式來納入所有子代的基因資訊,來偵測是否存在著罕見變異的關聯性分析。這裡所提出的不確定性編碼矩陣可以同時考慮單倍體組成狀態不明,以及單倍體傳遞狀態的不確定性,並且會使用一種以演化為基礎的資料分群法來降低維度。模擬分析與實際應用已經完成,相較於現有的分析工具,我這裡提出的方法有比較好的分析結果。以上這些演算法都將以R程式語言來完成,未來將會放在網路上提供研究者免費下載使用。 Compared with single marker analysis, marker-set analysis usually contains more biological interpretation and can provide more informative and powerful results. For instance, the haplotype with linkage-disequilibrium among SNP markers is one marker-set presentation. Such analysis, however, does not come free. There are difficulties need to be overcome. For example, the determination of haplotype phase and the large number of haplotypes in the genetic region of interest may cause problem in statistical inference. In addition, for family study designs, there exists the uncertainty of transmission/non-transmission status from parents to affected offspring. In this thesis, I will construct an uncertainty-coding matrix of marker-set based on collected genotype data, and apply to association studies with family design for two stages. First, I will base on trios design to conduct the transmission/non-transmission haplotype and employ the Bayesian conditional logistic regression. Second I use Bayesian generalized mixed effect model to incorporate the whole marker-set information of the other offspring in family. Such design matrix can cope with the phase uncertainty and the transmission uncertainty among family studies. Furthermore, an evolutionary-based clustering method can avoid the curse of dimensionality. The simulation studies and real data applications are presented and compared with other tools. These proposed methods are implemented in R and will be available for free download in the future. |
URI: | http://ntur.lib.ntu.edu.tw//handle/246246/250274 |
顯示於: | 流行病學與預防醫學研究所 |
檔案 | 描述 | 大小 | 格式 | |
---|---|---|---|---|
ntu-101-D95842002-1.pdf | 23.32 kB | Adobe PDF | 檢視/開啟 |
在 IR 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。