在乳癌中全基因組鑑定調節基因

Abstract

隨著生物晶片技術的純熟，利用統計檢驗來分析差異表現基因是目前最常用的策略。近年來，已有大量的研究去檢測重要的差異表現基因，透過其基因調控、可讓我們更加了解癌症分子的作用機制及提供有效預後診斷。然而, 此方法找出的基因調控關係、往往只存在於某種特定情況，是一種靜態的表徵, 卻忽略了事實上基因在細胞間調控關係是一個動態的過程。因此，動態基因調控網路分析可提供更好的詮釋，讓我們理解複雜的生物現象。在癌症裡，動態基因調控關係會被某些重要的調節基因(modulator gene)所影響，特別是，當調節基因高度或低度表現時。已有報導指出, 致癌基因或是腫瘤抑制基因會藉由扮演調節基因的角色來影響基因的功能，例如乳癌中的雌激素受體( estrogen receptor, ER)。目前已有幾數種分析單一調節基因調控網路的計算方法。然而系統化且全基因組檢測重要調節基因卻是前所未有的。 在本篇的研究中，我們利用了一個Fisher 轉換的統計模型，藉由分析乳癌病人的基因表現量，整體性地鑑定關鍵調節基因。我們設計了兩個參數Con及DGC來衡量調節基因調控基因網路的強弱。ESR1是一個已被廣泛研究的調節基因，在本篇分析中，兩個參數也顯示ESR1確實具有顯著能力調節基因網路，足以證明該方法確實可以鑑定具有代表性的調節基因。因此，我們找出了237個關鍵調節基因，其中51個調節基因具有比ESR1更強的能力調節基因網路。為了驗證調節基因的可靠性，我們採用歸一化坎培拉距離(Canberra distance)來衡量兩個參數在跨乳癌資料組中調節基因能力排名的相似度。 為了探討237個關鍵調節基因所影響的生物功能，我們建立了調節基因 調控的基因網路。隨著調節基因狀態的不同，我們發現新的樞紐基因(hub gene)調 控基因的關係與其相關的生物功能。此外，我們還找出15個關鍵調節基因會因其狀態不同而影響病人的存活率，並可在跨乳癌資料組中得到驗證。 在本研究中，我們綜合性地分析乳癌中的每個調控基因的基因網路。期盼藉此可以加深我們對調節基因網路的認知及系統性的影響，幫助我們在癌症中發現複雜的基因調控機制。

With well-developed technology in microarray, using statistical tests to analyze microarray data is one of the most common strategies to identify differentially expressed genes. In recent years, tremendous efforts have been made to discover gene regulation among differentially expressed genes to provide insights into the molecular mechanisms and effective prognosis in cancers. However, such gene interactions have focused on certain static cellular conditions, ignoring the fact that interaction among genes in cells is a dynamic process. Dynamic gene regulation networks analysis provide better interpretation and understanding for complex biological phenomena. In cancers, the dynamic gene regulation can be modulated by some key modulator genes. That is some inter-gene regulation can be strengthened (or weakened) specifically, when certain modulator genes are over- or under-expressed. Some oncogenes or tumor suppressors have been reported to perform their function through acting as modulator genes, such as estrogen receptor (ER) in breast cancers. Several computational methods were developed to dissect the modulated regulatory networks under single modulator gene. However, systematic and genome-wide screening of novel modulator genes is not previously explored. In the study we developed a statistical model based on Fisher transformation analysis pipeline for comprehensively identifying key modulator genes from genome-wide expression profiling of breast cancer. We designed two parameters, connectivity (Con), and degree of genome-wide change (DGC), to measure the modulation strength. In our data, ESR1, the most well-studied modulator gene, exhibited significant power of modulation in both parameters, revealing that the proposed model was sensitive and capable of identifying modulators. Therefore, we identified 237 key modulator genes, 51 of them with higher power than ESR1. In order to statistically evaluate the robustness of the proposed model, we employed the normalized Canberra distance to measure similarity among the ranked lists of modulator genes in the discovery dataset and three independent validation datasets. For biological interpretation of these 237 identified key modulators, regulatory networks governed by these modulators were constructed. We identified novel regulation and functional interaction of hub gene under specific modulator gene status. Furthermore, 15 survival-associated modulator genes, which dependent on modulator gene status were identified and verified in the validation datasets. In summary, we provided a comprehensive analysis of measuring gene regulation by each modulator gene and constructed modulated gene regulation networks in breast cancer. We expect that the modulator identification of the gene interactions will further enhance our understanding in gene interaction and systemic influence to discover complex genetic regulations in cancers.