- 學位論文
Affymetrix 微陣列晶片配對實驗設計之基因表現與類型比較
Paired and Unpaired Design in Class Comparisons for Affymetrix Microarray
摘要
微陣列基因晶片是近代生物資訊重要的研究工具之一，而且微陣列晶片表現資料分析也是近代生物統計重要的議題之一。微陣列基因晶片常見的主要類型有：點狀式互補核甘酸 (spotted cDNA) 微陣列晶片與寡核甘酸 (oligonucleotide) 微陣列晶片等，後者以 Affymetrix 生技公司為代表。目前已有許多研究討論關於利用點狀式cDNA 晶片進行研究之實驗設計，且點狀式 cDNA 晶片之實驗設計大多利用配對控制或參照樣本；而對於利用寡核甘酸晶片之實驗設計則較少利用配對設計，其中一個原因是 Affymetrix 微陣列晶片利用PM-MM (perfect-match, mis-match) 探針組設計，提高特異性與敏感度。有一些生物資訊研究人員將配對樣本實驗設計應用在Affymetrix 微陣列晶片，但對於利用Affymetrix微陣列晶片之配對實驗設計之優缺點，則較少為統計人員討論。 本研究使用一個配對實驗設計Affymetrix微陣列晶片之資料，收集 26 位肺癌病患，26 組配對腫瘤組織與正常組織，共 52 片晶片，探討配對樣本與獨立樣本在類型比較中，檢定腫瘤與正常基因表現是否有差異，並且使用拔靴法探查此差異之變異性。 研究結果顯示，Affymetrix 微陣列晶片配對實驗設計，個體內基因之相關係數變異性大，且並未能有效地降低個體間基因之間的相關性，在檢定兩組樣本基因表現是否有差異時，在配對與獨立樣本中，顯著表現基因數目的變異性大。
並列摘要
Microarray is one of the important research tools in modern bioinformatics. And gene expression data analysis is also one of the important topics in modern biostatistics. Two major microarray platforms exist: cDNA microarray and oligonucleotide microarray; and the later is represented by Affymetrix. Many literatures have discussed about the spotted cDNA experiment designs, and the spotted cDNA experiment designs often adapt paired design that include paired controls or reference samples. However, oligonucleotide microarray experimental designs rarely adapt paired design. One of the reasons is that Affymetrix microarray is designed with PM-MM (perfect-match, mis-match) probe cells such that it increases specificity and sensitivity. Some bioinformatics researchers apply paired experiment design on Affymetrix microarrays. However, few researches report the benefits of the paired experiment design on Affymetrix microarray. In this empirical study, we use an Affymetrix microarray gene expression data set that includes 26-pair chips. This data set came from a lung cancer survival study that included 26 subjects. Paired tumor and adjacent normal tissues were sampled. Class comparisons between tumor and normal tissues were analyzed to search significant differential expression genes. The variation of the significant genes is evaluated by bootstrap method. The result shows that variations of pairwise gene correlation are large. And paired design can not efficiently increase within-gene correlations and reduce between-gene correlations. The variation of the significant genes is also large that can not be reduced by paired design. More empirical data are necessary to assess the benefit of paired design for Affymetrix microarray