運用核糖核酸測序實驗資料以偵測有差異表現量之基因

Abstract

已知核糖核酸測序實驗(RNA-Seq)較微陣列實驗(microarray)具備多項優勢，如可避免雜交過程所帶來的雜訊，故準確度較高，以及不受序列為已知的需求限制等。預期隨著實驗成本逐漸降低，核糖核酸測序實驗將逐漸受到歡迎並且被廣泛運用。本研究將針對運用該類實驗資料以偵測顯著差異基因的問題提出統計分析方法。已知所獲得的基因資料為計數型態(count data)，在考慮其過度離散(over-dispersion)性質下，我們將以負二項(Negative Binomial)分配來配適資料。為了降低計算難度，將採用最大擬概似函數(maximum pseudo likelihood)估計法。在基因的差異顯著性檢定上，我們則採用外顯類組(phenotypic group)均數差之Wald檢定統計量。本研究透過電腦模擬以驗證所提出的方法，並且運用此方法在數組實際資料上以評估其實用性。

Recently, the RNA-Seq experiment is developed for a high-throughput DNA sequencing method for mapping and quantifying the transcriptomes. The gene expression level obtained from a RNA-Seq experiment is of the count data type and is often fitted by a Negative-Binomial distribution to account for the over-dispersion. To identify the differentially expressed genes with a binary phenotypic response, we aim to develop a statistical test for comparing the means of two Negative-Binomial distributions. To ease the computational complexity, a maximum pseudo likelihood estimation (MPLE) is considered and the corresponding Wald’s test statistic is subsequently employed. Simulation studies are performed to justify the adequacy of the proposed test in comparison with existing methods. The applicability of the proposed method is demonstrated via the data analysis of some real example data sets.