臺灣博碩士論文加值系統

在分子生物學的領域上，利用統計方法分析基因晶片的資料已成為一種趨勢。若能因此發掘出造成疾病的關鍵基因，對人類會有重要的貢獻。本篇文章中，基於致病基因會在生病的群體中有異常的表現，我們提供一些統計方法能在眾多基因中找出可能致病的關鍵基因。這些方法包含了 WORT、 WOS、PGM、TGM、QGM，以及 BRP。我們也將這些方法與過去曾經被發表的 T-statistic、OS、OR 以及COPA等四個方法做比較。

It's a trend to use statistical methods in medical science. If the genes which cause the diseases could be found, it might be helpful to nowadays medical field. In this article, we proposed several methods to find the probable influential genes which are over- or down-expressed in some but not all samples in a disease group. Those methods include WORT (weight outlier robust t-statistic), WOS (weight outlier sum), PGM (the MLE of probability of Gaussian mixture model), TGM (T-statistic of Gaussian mixture model), QGM(Quantile of Gaussian mixture model), and Bayesian Rule P-value(BRP). Also we will compare those methods with four methods (t-statistic, OS, ORT, COPA) which have been proposed and published for detecting differentially expressed genes. Those new methods include improvements of ORT and OS methods, four methods related to Gaussian mixture model and Bayesian method.

1 Introduction 1
2 Statistical Methods 2
2.1 methods review . . . . . . . . . . . . . . . . . . . 2
2.1.1 T-statistic . . . . . . . . . . . . . . . . . . . .2
2.1.2 The Outlier Sum . . . .. . . . . . . . . . . . . . 2
2.1.3 The Outlier Robust T-statistic . . . . . . . . . . 3
2.1.4 Cancer Outlier Profile Analysis. . . . . . . . . . 4
2.2 New methods . . . . . . . . . . . . . . . . . . . . 5
2.2.1 The Weighted OS . . . . . . . . . . . . . . . . . 5
2.2.2 The Weighted ORT . . . . . . . . . . . . . . . . . 5
2.2.3 Methods related Gaussian mixture model . . . . . . 6
2.2.4 Bayesian Rule P-value . . . . . . . . . . . . . . 7
3 Simulation Study 9
3.1 Comparison by mean, median, Q1, and Q3 . . . . . . . 9
3.2 Comparison by the true/false-positive rates plots . 10
4 Real Data 20
5 Appendix 21
5.1 T-statistics . . . . . . . . . . . . . . . . . . . 21
5.2 COPA . . .. . . . . . . . . . . . . . . . . . . . . 24

1.Bickel, P.J. and Levina, E. (2003) Some theory for Fisher's Linear Discriminant function, ``naive Bayes'', and some alternatives when there are many more variables than observations.
2.Fan, J. and Fan, Y. (2007). High Dimensional Class Using Features Annealed Independence Rules.
3.Tibshirani, R. and Hastie, T. (2007). Outlier sums for differential gene expression analysis. Biostatistics, 8, 1,pp.2-8.
4.Wu, B. (2007). Cancer outlier differential gene expression detection. Biostatistics, 8, 3, pp. 566-575.
5.P. Baldi and A.D. Long. (2001). A Bayesian framewoek for the analysis of microarray expression date: regularized t-test and statistical inferences of gene change. Bioinformatics, 17(6):509-519, 2001.
6.Erik, K., Anders, S, Mats, R., and Olle, N. (2007). Weighted analysis of general microarray experiments.