微陣列實驗是在近年來最熱門生物技術。藉由此高效率的技術可快速比較大量的基因表現,並可進一步了解基因和疾病的顯著性。微陣列實驗包含了許多的步驟,容易產生一些非生物上的變異,造成了不正確的分析解讀。為了增加分析微陣列結果的正確性,所以在資料的前置處理工作就變得十分重要。對於Affymetrix Oligonucleotide 微陣列的前置處理方法中,分位數常規化方法是最廣為使用的。在本研究中,我們使用主成分分析來提出加權分位數常規化法。本文使用Affymetrix 公司所提供的資料之HGU133 和GEO 網站中的數個實際資料,將RMA 和 GCRMA 所提的前置處理方法和加權分位數常規化法做比較。並利用Affycomp II 網站中的指標和基因差異表現分析來評估方法的優劣。分位數常規化方法和加權分位數常規化方法, 雖然以不同的出發點有所差異, 可是這二個方法在前置處理的能力上卻擁有許多相似之處。但若同組資料本身變異程度較大,在基因差異表現分析上略有所不同。 In recent years, the microarray experiment has become the most popular biotechnology to study the gene expression. The expression levels of thousands of genes are simultaneously measured to investigate the association of certain treatments, diseases, and genes. In order to remove the impact of nonbiological variations and systematic bias presents in such high-throughput data, the pre-processing is an essential and important step in microarray data analysis. Among these, RMA (robust multiarray averaging) and GCRMA are the most widely used pre-processing methods for Affymetrix GeneChip data. Both methods use the quantile normalization for the normalization step. In this study, we proposed a weighted quantile normalization using the principal component analysis. The standard HGU133 dataset from Affymetrix and the other 14 datasets from GEO website were employed to compare RMA, GCRMA and their weighted versions. The evaluation was reported based on the indices of Affycomp II and the significance analysis of microarrays (SAM). The finding suggests the differential expressed genes found by the weighted quantile normalization were slightly different with those obtained from the classical method if the input data possesses large variations.