Evaluation of Missing Value Estimation for Microarray Data

Danh V Nguyen,Raymond J Carroll,Naisyin Wang

doi:10.6339/jds.2004.02(4).170

Abstract

Publisher: School of Statistics, Renmin University of China, Journal: Journal of Data Science, Title: Evaluation of Missing Value Estimation for Microarray Data, Authors: Danh V. Nguyen, Naisyin Wang, Raymond J. Carroll

Highlights

Introduction and BackgroundDNA microarrays, designed to monitor mRNA expression levels of thousands of genes in concert, are used to investigate various biological processes
The fifth data set used to evaluate estimation accuracy is a cDNA data set consisting of 7 breast cancer (BC) samples with mutation in the BRCA1 gene, 8 with mutation in the BRCA2 gene, and 7 sporadic cases with neither mutations detected (Hedenfalk et al, 2001)
The complete expression matrix of the combined 22 BC samples consists of 3,226 cDNAs

Summary

Introduction

Introduction and BackgroundDNA microarrays, designed to monitor mRNA expression levels of thousands of genes in concert, are used to investigate various biological processes. Gene expression data obtained from microarray experiments, like other experimental data, often contain missing values (MVs). Some data analysis methods applied to gene expression data, including some classification and model-based clustering techniques, are not equipped to handle missing data. J. Carroll expression matrix, the primary approaches to missing data include (1) removing data points with MVs before the analysis or (2) estimating the MVs and proceeding to the analysis. Approach (2), estimating the MVs before analysis, is less common and only naive methods, such as replacing MVs with zeros or the sample means, have been used. One of the earliest use of a more sophisticated MV estimation method is by Dudoit, Fridlyand and Speed (2002), where the method of K-nearest neighbors (KNN) was used to estimate MVs before applying various classification methods

Methods

Results

Discussion

Conclusion