在自足檢定下進行含拓墣意涵之生物路徑相關性研究

Abstract

現今研究認為許多複雜疾病可能由多個效果中等的基因一起作用，共同影響疾病的發生；因此，許多研究利用單一標記檢定找出與疾病有關的基因，進行以基因集合為單位的基因集合分析(Gene-set analysis)；其中包含針對生物路徑(pathway)的基因集合分析。然而，生物路徑為一種帶有生物功能與機制描述的基因集合，生物路徑內的基因之間具有生物上的關係，但這些訊息並未被納入過去的基因集合分析。本論文提出一個結合單一標記檢定、皮爾森相關係數及生物路徑拓樸概念的方法，來偵測生物路徑與疾病間的相關性。並利用模擬資料與一高程度乳腺管原位癌(High-Grade Ductal Carcinoma in Situ)的次世代定序資料比較了六種不同的基因集合分析統計檢定。初步結果顯示，當基因間關係為獨立時，本研究方法效果並沒有特別突出，而當模擬的基因資料彼此有相關性時且生物路徑中同時含有致病及保護基因時，本研究方法效果較佳；而且，觀察幾個與乳癌有關的生物路徑內基因彼此的相關係數確實較高，模擬具相關性的情境也較符合真實資料。最後，本研究也提出如何透過拓樸的概念，利用模擬資料及實際乳癌資料來評估生物路徑內每一基因與疾病的相關性。

Many complex diseases are associated with multiple genes, each with a mild effect, rather than a single gene with a strong effect. The gene-set analysis can examine if a collection of genes, such as a pathway, is associated with the disease of interest. However, most current methods consider genes in the same pathway as independent units. They did not consider the relationship between genes. In this thesis, we proposed a method which takes account of the results from single-marker tests, correlation between genes, and the topology of pathway in constructing the association test statistic. For performance evaluation, simulations and a high-grade ductal carcinoma in situ study were illustrated, and the proposed test was compared with other gene-set analyses. Our proposed method performs better than other methods when correlation exists between genes. When genes are independent, our method performs similarly to others. Finally, we comment on the detection of influential genes using the same rationale, and prioritize the importance of genes in the associated gene-set.