Design and Analysis of DNA Microarray Investigations (Statistics for Biology and Health)

This Book Is Targeted To Biologists With Limited Statistical Background And To Statisticians And Computer Scientists Interested In Being Effective Collaborators On Multi-disciplinary Dna Microarray Projects. State-of-the-art Analysis Methods Are Presented With Minimal Mathematical Notation And A Focus On Concepts. This Book Is Unique Because It Is Authored By Statisticians At The National Cancer Institute Who Are Actively Involved In The Application Of Microarray Technology. Many Laboratories Are Not Equipped To Effectively Design And Analyze Studies That Take Advantage Of The Promise Of Microarrays. Many Of The Software Packages Available To Biologists Were Developed Without Involvement Of Statisticians Experienced In Such Studies And Contain Tools That May Not Be Optimal For Particular Applications. This Book Provides A Sound Preparation For Designing Microarray Studies That Have Clear Objectives, And For Selecting Analysis Tools And Strategies That Provide Clear And Valid Answers. The Book Offers An In Depth Understanding Of The Design And Analysis Of Experiments Utilizing Microarrays And Should Benefit Scientists Regardless Of What Software Packages They Prefer. In Order To Provide All Readers With Hands On Experience In Data Analysis, It Includes An Appendix Tutorial On The Use Of Brb-arraytools And Step By Step Analyses Of Several Major Datasets Using This Software Which Is Freely Available From The National Cancer Institute For Non-commercial Use. The Authors Are Current Or Former Members Of The Biometric Research Branch At The National Cancer Institute. They Have Collaborated On Major Biomedical Studies Utilizing Microarrays And In The Development Of Statistical Methodology For The Design And Analysis Of Microarray Investigations. Dr. Simon, Chief Of The Branch, Is Also The Architect Of Brb-arraytools. Introduction -- Dna Microarray Technology -- Design Of Dna Microarray Experiments -- Image Analysis -- Quality Control -- Array Normalization -- Class Comparison -- Class Prediction --class Discovery -- A: Basic Biology Of Gene Expression -- B: Description Of Gene Expression Datasets Used As Examples -- C: Brb-array Tools. Richard M. Simon ... [et Al.]. Includes Bibliographical References (p. [185]-194) And Index. Contents......Page 8 Acknowledgments......Page 6 1 Introduction......Page 12 2.2 Measuring Label Intensity......Page 16 2.3 Labeling Methods......Page 17 2.4 Printed Microarrays......Page 18 2.5 Affymetrix GeneChipTM Arrays......Page 20 2.6 Other Microarray Platforms......Page 21 3.1 Introduction......Page 22 3.2.1 Class Comparison......Page 23 3.3 Comparing Two RNA Samples......Page 24 3.4 Sources of Variation and Levels of Replication......Page 25 3.5 Pooling of Samples......Page 27 3.6.1 The Reference Design......Page 28 3.6.2 The Balanced Block Design......Page 30 3.6.3 The Loop Design......Page 31 3.7 Reverse Labeling (Dye Swap)......Page 32 3.8 Number of Biological Replicates Needed......Page 34 4.1 Image Generation......Page 40 4.2.2 Gridding......Page 41 4.2.3 Segmentation......Page 42 4.2.4 Foreground Intensity Extraction......Page 43 4.2.5 Background Correction......Page 44 4.2.6 Image Output File......Page 45 4.3 Image Analysis for Affymetrix GeneChipTM......Page 46 5.1 Introduction......Page 50 5.2.2 Spots Flagged at Image Analysis......Page 51 5.2.3 Spot Size......Page 52 5.2.4 Weak Signal......Page 53 5.2.5 Large Relative Background Intensity......Page 54 5.3 Gene Level Quality Control for Two-Color Arrays......Page 55 5.3.2 Probe Quality Control Based on Duplicate Spots......Page 56 5.3.3 Low Variance Genes......Page 57 5.4 Array-Level Quality Control for Two-Color Arrays......Page 58 5.5 Quality Control for GeneChipTM Arrays......Page 59 5.6 Data Imputation......Page 61 6.2.1 Biologically Defined Housekeeping Genes......Page 64 6.2.2 Spiked Controls......Page 65 6.3 Normalization Methods for Two-Color Arrays......Page 66 6.3.1 Linear or Global Normalization......Page 67 6.3.2 Intensity-Based Normalization......Page 68 6.3.3 Location-Based Normalization......Page 70 6.4.1 Linear or Global Normalization......Page 72 6.4.2 Intensity-Based Normalization......Page 73 7.1 Introduction......Page 76 7.2 Examining Whether a Single Gene is Differentially Expressed Between Classes......Page 77 7.2.1 t-Test......Page 78 7.2.2 Permutation Tests......Page 79 7.2.3 More Than Two Classes......Page 82 7.2.4 Paired-Specimen Data......Page 84 7.3 Identifying Which Genes Are Differentially Expressed Between Classes......Page 86 7.3.1 Controlling for No False Positives......Page 87 7.3.2 Controlling the Number of False Positives......Page 91 7.3.3 Controlling the False Discovery Proportion......Page 92 7.4 Experiments with Very Few Specimens from Each Class......Page 95 7.5 Global Tests of Gene Expression Differences Between Classes......Page 97 7.6 Experiments with a Single Specimen from Each Class......Page 99 7.7 Regression Model Analysis; Generalizations of Class Comparison......Page 101 7.8 Evaluating Associations of Gene Expression to Survival......Page 102 7.9 Models for Nonreference Designs on Dual-Label Arrays......Page 103 8.1 Introduction......Page 106 8.2 Feature Selection......Page 108 8.3.2 Discriminant Analysis......Page 109 8.3.3 Variants of Diagonal Linear Discriminant Analysis......Page 112 8.3.4 Nearest Neighbor Classification......Page 114 8.3.5 Classification Trees......Page 115 8.3.6 Support Vector Machines......Page 117 8.3.7 Comparison of Methods......Page 118 8.4.1 Bias of the Re-Substitution Estimate......Page 119 8.4.2 Cross-Validation and Bootstrap Estimates of Error Rate......Page 121 8.4.3 Reporting Error Rates......Page 123 8.4.5 Validation Dataset......Page 124 8.5 Example......Page 125 8.6 Prognostic Prediction......Page 129 9.1 Introduction......Page 132 9.2 Similarity and Distance Metrics......Page 133 9.3.1 Classical Multidimensional Scaling......Page 136 9.4.1 Hierarchical Clustering......Page 142 9.4.2 k-Means Clustering......Page 149 9.4.3 Self-Organizing Maps......Page 153 9.4.4 Other Clustering Procedures......Page 156 9.5 Assessing the Validity of Clusters......Page 157 9.5.1 Global Tests of Clustering......Page 159 9.5.2 Estimating the Number of Clusters......Page 161 9.5.3 Assessing Reproduciblity of Individual Clusters......Page 163 A.1 Introduction......Page 168 B.2 Bittner Melanoma Data......Page 176 B.4 Perou Breast Data......Page 177 B.5 Tamayo HL-60 Data......Page 178 B.6 Hedenfalk Breast Cancer Data......Page 179 C.1 Software Description......Page 180 C.2 Analysis of Bittner Melanoma Data......Page 182 C.3 Analysis of Perou Breast Cancer Chemotherapy Data......Page 189 C.4 Analysis of Hedenfalk Breast Cancer Data......Page 193 References......Page 196 C......Page 206 G......Page 207 N......Page 208 R......Page 209 W......Page 210 DNA microarrays are an important technology for studying gene expression. With a single hybridization, the level of expression of thousands of genes, or even an entire genome, can be estimated for a sample of cells. Consequently, manylaboratoriesareattemptingtoutilizeDNAmicroarraysintheirresearch. Whereaslaboratoriesarewellpreparedtoaddressthesigni?cantexperimental challenges in obtaining reproducible data from this RNA-based assay, inv- tigators are less prepared to analyze the large volumes of data produced by DNA microarrays. Although many software packages have been developed for the analysis of DNA microarray data, software alone is insu?cient. One needs knowledge aboutthevariousaspectsofdataanalysisinordertoselectandutilizesoftware e?ectively. There is a plethora of analysis methods being published and it is di?cult for biologists to determine which methods are valid and appropriate for their problems. Many scientists have learned that software is not an adequate substitute for biostatistical knowledge and seek statistical collaborators. Unfortunately, there is presently a shortage of statisticians who are available and knowled- able about DNA microarrays. For statisticians to be e?ective collaborators in anyarea,theymustinvestthetimetounderstandthesubjectmatterareaand become familiar with the literature so that they can ask the right questions and identify the key issues. Our objectives in this book are twofold: to provide scientists with infor- tion about the design and analysis of studies using DNA microarrays that will enable them to plan and analyze their own studies or to work with statistical collaborators e?ectively, and to aid statistical and computational scientists wishing to develop expertise in this area. This book discusses the analysis of gene expression profile data from DNA micorarray studies and the design of such studies. The book will address design and analysis issues for both of the major classes of DNA microarrays, the cDNA microarrays and the oligonucleotide arrays. DNA microarrays are a new technology that is revolutionizing biological and biomedical research. Most biologists are trying to analyze their own data using a variety of commercial and public domain software. This book will provide an authoritative review of the methods available and present it in a manner that is intelligible to biologists. The analysis of gene expression profile data from DNA micorarray studies are discussed in this book. It provides a review of available methods and presents it in a manner that is intelligible to biologists. It offers an understanding of the design and analysis of experiments utilizing microarrays to benefit scientists. It includes an Appendix tutorial on the use of BRB-ArrayTools and step by step analyses of several major datasets using this software which is available from the National Cancer Institute.