- 學位論文
利用基因群層級之存活分析方法鑑定肺癌之穩定預後基因標記
Identification of Common Prognostic Gene Expression Signature in Lung Cancer with Gene-Set Level Survival Analysis
摘要
近年來，肺癌在世界上高居癌症死亡原因的前幾位。其中非小細胞肺癌(non-small cell lung cancer)佔了大部分的肺癌案例。到目前為止，仍舊有許多我們所不了解的異質性存在於肺癌的病患當中，因而間接造成了長期以來肺癌居高不下的死亡率。復發風險至今還是難以預測，而且即使是在擁有相近臨床特徵的病人當中，最終的結果依然存在很大的變異性。因此，為了改善不同子型肺癌病人的預後診斷及管理，對於尋找出新的分子標記來輔助現今的臨床特徵，產生了急迫的需求。 大部分用來尋找預後標記的方法都是基於單一基因的統計分析，這種方式造成了幾個主要的問題。不同研究所找出來的預後標記彼此之間的重疊性往往很低，而當這些標記被應用到許多不同的資料組上的時候，往往也缺乏穩定的表現。另一方面，存在這些預後標記當中的生物意義，以及他們彼此間互動的關係，常常缺乏合理的解釋。對於為何這些標記會影響病人最終的結果，也因而相當難以闡述。 在本研究當中，我們提出了一個跨平台、基於基因群層級的存活分析方法。此方法結合了Cox正比例風險迴歸模型(Cox proportional hazard regression model)與基因群富集分析(gene set enrichment analysis, GSEA)。透過此方法，基於存在於基因群定義當中的知識，生物意義在一開始就被納入。同時，因為預後標記彼此間功能上的互動關係增多了，鑑定出來的預後標記在跨資料組的穩定度也獲得了提昇。藉助這些穩定的預後標記的幫助，病人的預後診斷及存活率在未來將能夠獲得顯著的提昇。
並列摘要
Lung cancer has been the leading cause of cancer death worldwide in recent years and non-small cell lung cancer (NSCLC) accounts for most cases of lung cancer. There is still great heterogeneity poorly understood in lung cancer, which accounts for poor survival rate. Risk of recurrence of lung cancer is not predictable and a large variability of disease outcome has been observed for patients with similar clinical features. Therefore, there is an urgent need to discover new molecular signatures that could incorporate with current clinical features to improve the prognosis and management of patients within each subtype of NSCLC. Most approaches for finding prognostic signatures were based on individual gene testing, which caused several major problems. First, there were only limited overlaps between prognostic signatures identified from different studies and these signatures demonstrated a lack of robustness when applying to multiple datasets. Moreover, it was difficult to elucidate the biological mechanisms underlining those prognostic signatures in connection to clinical outcomes. In this study, a cross-platform gene-set level approach integrating the cox proportional hazard regression model with gene set enrichment analysis (GSEA) was proposed. Biological meaning was included at first based on prior knowledge in gene-set definition. Robustness of signatures was also strengthened in companion with more functional interactions among these signatures. With the help of robust prognostic signatures, prognosis and survival of patients could be improved a lot in the future.