相對 R 平方方法理論性質的探討

Abstract

相對R-squared 方法(RRSM) 是近幾年來被提出來尋找miRNA 的標靶基因或致病基因的有效方法。傳統上，生物學家在分析microarray data來尋找miRNA標靶基因時大多應用相關性分析，但在實際的例子發現相關性分析無法得到與生物實驗相同的結果。而其它被提出的統計方法也無法在一些實驗數據上得到好的結果。所以，Wang and Li (2009) 利用迴歸模式去建立一個相對R-squared 方法，而此方法在利用microarray data 來尋找miRNA 的標靶基因的分析上比相關分析和其它被提出的方法有更好的表現。雖然相對R-squared 方法在實際應用已有好的結果，但它的理論性質還沒有被仔細的探討。在此研究計劃中，我們著力於相對R-squared 方法的理論性質的探討並利用統計推論方法來推導它的統計性質如區間估計等。

The relative R-squared method (RRSM) has recently been proposed to discover miRNA interactions. Traditionally, the statistical method for obtaining the miRNA targets from the microarray expression data is based on correlation analysis. However, the experimental results reveal that the correlation analysis cannot lead to very accurate results resulting in a large false discovery rate. Therefore, Wang and Li (2009) establish a relative squared method based on a regression model to find miRNA interactions from microarray data, which is shown to outperform the correlation analysis and another existing method. Although RRSM is successful in selecting high-confidence miRNA targets from the microarray data, its theoretical properties have not been deeply explored. In this study, we will investigate and derive the theoretical properties for RRSM.