【論】Rachidi,Cystic Fibrosis Gene Expression Data～

Thami Rachidi & Krzysztof Cios
Cystic Fibrosis Gene Expression Data Analysis

・DNAマイクロアレイデータを使い、4つのデータ正規化法の性能を比較。2種の検定法でそれぞれ遺伝子抽出し、FDRで評価。
・データ：嚢胞性肺繊維症（CF:Cystic Fibrosis）と正常（control）のサンプルが各9個。12,225遺伝子。Affymetrix製。
・4つのデータ正規化法
1.Affymetrix Microarray Suite 5.0(MAS 5)
2.Robust multichip analysis(RMA)
3.DNA-Chip (dChip) based on perfect match only
4.dChip based on perfect match/mismatch difference
・2つの検定法
1.t-test (従来法)
2.KS-test（Kolmogoro-Smirnov test）（non-parametricな方法）

・サンプル数が足りず、期待するような結果がでなかった。というオチなので論文にはならなかったレポート。

・「The problem with the t-test approach is that it assumes that the distribution of the mean being tested is normal. This is not true for many gene expression data.」
・「It appears that for all normalization algorithms, t-test gives better results than the non-parametric Kolmogoro-Smirnov test. This was expected since more replicates in the CF data are needed in order to achive good results with most of the non-parametric methods.」
・「The choice of normalization seems to greatly affect the differentially expressed genes across the conditions.」