錯誤發現率控制方法變異之量度

Characterizing the Variability of False Discovery Rate Control

林奕廷(Yi-Ting Lin)

指導教授 ： 李文宗

國立臺灣大學/公共衛生學院/流行病學與預防醫學研究所/碩士(2013年)

https://doi.org/10.6342/NTU.2013.10682

摘要

在進行DNA微陣列基因晶片資料分析時，由於檢定次數龐大，會遇到多重檢定問題。控制錯誤發現率是一個新的思維，可以有較Bonferroni校正更佳的檢定力。然而研究者忽略了錯誤發現率控制方法有三個層次的變異，進而做出錯誤的闡釋。本研究中，我們利用拔靴法估計局部錯誤發現率、錯誤發現率控制數值與偽陽性個數比例等三個層次的變異。作者使用大腸癌基因表現實際資料與眼屈光不正GWAS實際資料，呈現三個層次變異大小。兩筆真實資料皆顯示明顯的變異。在顯著的基因中，局部錯誤發現率、q-值與偽陽性個數比例的拔靴法標準誤最大分別可達0.0891、0.0273與0.0828。我們呼籲在使用錯誤發現率控制方法時，不應只列出錯誤發現率控制數值或q-值，還需呈現錯誤發現率控制方法有三個層次的變異，才可確保每個部分都會有正確的闡釋。

關鍵字

錯誤發現率控制方法 ； 拔靴法 ； 變異 ； 局部錯誤發現率 ； 偽陽性個數比例

並列摘要

The problem of multiple hypothesis testing arises during the analysis of DNA microarray data owing to numerous statistical tests. Controlling the false discovery rate (FDR) is a revolutionary idea with higher power than traditional error control procedures. However, there are three levels of variation in FDR control, which result in incorrect interpretations if ignored by investigators. In this study, we estimate the variation of local FDR, FDR control value, and false discovery proportion by bootstrap. A colon cancer gene-expression data and a visual refractive errors GWAS data will be analyzed as demonstration. Each data shows apparent variation. Among significant genes, the largest bootstrap standard error of local FDR, overall q-value and false discovery proportion are 0.0891, 0.0273 and 0.0828 respectively. Therefore, we urge the importance of reporting three levels of variation in FDR control to ensure correct interpretations.

並列關鍵字

false discovery rate control ； bootstrap ； variation ； local FDR ； false discovery proportion

參考文獻

1. Pounds, S.B., Estimation and control of multiple testing error rates for microarray studies. Brief Bioinform, 2006. 7(1): p. 25-36.

2. Benjamini, Y. and Y. Hochberg, Controlling the False Discovery Rate - a Practical and Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society Series B-Methodological, 1995. 57(1): p. 289-300.

3. Storey, J.D. and R. Tibshirani, Statistical significance for genomewide studies. Proc Natl Acad Sci U S A, 2003. 100(16): p. 9440-5.

4. Alon, U., et al., Broad patterns of gene expression revealed by clustering analysis of tumor and normal colon tissues probed by oligonucleotide arrays. Proc Natl Acad Sci U S A, 1999. 96(12): p. 6745-50.

5. Stambolian, D., et al., Meta-analysis of genome-wide association studies in five cohorts reveals common variants in RBFOX1, a regulator of tissue-specific splicing, associated with refractive error. Hum Mol Genet, 2013. 22(13): p. 2754-64.