即時聚合酶鏈鎖反應 (real-time polymerase chain reaction) 在基因檢測上是相當普遍的技術,該技術因可即時監測反應過程並對目標基因做定量分析,所以具更高的敏感性。即時聚合酶鏈鎖反應的基因表現定量方法可分為絕對定量與相對定量,由於絕對定量較耗費成本,因此較常使用相對定量。使用相對定量時,一般都以看家基因 (housekeeping gene) 當作內部控制基因來校正目標基因的表現量,以消除各樣本之間非生物性所產生的變異,但由於看家基因在不同的實驗條件下其基因的表現量不穩定 (Andersen 等人, 2004 及 Dheda 等人, 2004),因此本論文討論高密度寡核苷酸晶片之正規化方法應用於即時聚合酶鏈鎖反應基因表現相對定量法校正之可行性。 考慮三種高密度寡核苷酸晶片常用的正規化方法,分別為尺度正規化 (scaling normalization; Affymetrix, 2002)、不變集正規化 (invariant set normalization; Li 及 Wong, 2001)、百分位正規化 (quantile normalization; Bolstad 等人, 2003),將其應用於即時聚合酶鏈鎖反應基因表現相對定量法之校正上。由於兩種資料特性差異很大,本論文使用大量模擬情境分析與四種衡量方式來評估這三種正規化方法運用在即時聚合酶鏈鎖反應相對定量法之適切性,並且利用實例資料來檢驗各正規化方法的適用情形。研究結果發現在即時聚合酶鏈鎖反應基因表現相對定量法之校正上,除了使用看家基因的校正方式外,尺度正規化也是一個不錯的校正方法。
The real-time PCR (real-time polymerase chain reaction) is a common technique for evaluating the gene expression. This technique can provide very sensitive and accurate results since it is monitored instantaneously and also performs a quantitative analysis for the target gene. It has become a widespread technique in analyzing gene expressions. There are two methods to quantify the real-time PCR gene expression, relative and absolute quantification. Owing to cost and available sources, the relative quantification is the more commonly used method. However, the relative quantification requires a housekeeping gene as an internal control gene to normalize the target gene expression. Andersen et al. (2004) and Dheda et al. (2004) pointed out the gene expression of housekeeping gene may be unstable not only due to the biological variation, but also different experimental conditions. Hence, we discuss the feasibility of implementing the normalization method for high density oligonucleotide array to the relative quantification in real-time PCR. Three common normalization methods for high density oligonucleotide array, the scaling normalization (Affymetrix, 2002), the invariant set normalization (Li and Wong, 2001) and the quantile normalization (Bolstad et al. 2003), are discussed. Owing to large differences in data characteristics, Monte Carlo simulations are used to evaluate the performance of these normalizations to the real-time PCR. Four indices are used to assess the performance. Furthermore, a real data is used to illustrate the feasibility of these normalizations to the real-time PCR. We find that instead of using the housekeeping gene, the scaling normalization is a good choice for relative quantification in real-time PCR.