關於微陣列資料的前景與背景之變異數分析模型

Abstract

微陣列 (microarray) 技術是一種能偵測訊息核糖核酸 (mRNA, messenger ribosnucleic acid) 不同表現的有利工具。因其能提供大量的資料，此技術在最近被廣泛的利用在生物與醫藥應用方面，但微陣列在生物上、實驗系統與隨機的變化使得分析微陣列的資料具有很大的挑戰性。本篇論文提出前景與背景之變異數分析模型 (foreground and background ANOVA models) 來分析微陣列資料中的生物、實驗與隨機的影響，利用加權最小平方法 (Weighted least squares estimation) 估計模型中的參數，並且採用重抽樣的方法，例如: 排列檢定(permutation tests)與拔靴法 (boostrap) ，得到信賴區間與顯著機率值(p-values)。本篇的結果與即時聚合鏈反應 (Real Time PCR) 和其他相關微陣列實驗的結果作比較。

Spotted cDNA microarray (i.e., microscopic array) is a powerful tool to detect differential expressed genes in mRNA. Because of its capacity of high throughput, it has been widely used in biological and medical applications recently. Due to the biological, systematic and random variations of microarrays, it is a big challenge to analyze microarray data. In this paper we propose the foreground and background ANOVA models to explore the effects of biological, systematic and random variations from the microarray data. Weighted least squares estimation by the approach of coordinatewise descent is proposed to estimate the parameters. The confidence intervals and p-values are obtained by the resampling methods, including permutation tests and bootstrap methods. The results are compared with the findings by Real Time PCR and other microarray experiments.