Statistics for Microarrays Design, Analysis and Inference

by Wit, Ernst; McClure, John

• ISBN13: 9780470849934
• ISBN10: 0470849932
• Edition: 1st
• Format: Hardcover
• Copyright: 2004-07-23
• Publisher: WILEY

Summary

Interest in microarrays has increased considerably in the last ten years. This increase in the use of microarray technology has led to the need for good standards of microarray experimental notation, data representation, and the introduction of standard experimental controls, as well as standard data normalization and analysis techniques. Statistics for Microarrays: Design, Analysis and Inference is the first book that presents a coherent and systematic overview of statistical methods in all stages in the process of analysing microarray data - from getting good data to obtaining meaningful results. Provides an overview of statistics for microarrays, including experimental design, data preparation, image analysis, normalization, quality control, and statistical inference. Features many examples throughout using real data from microarray experiments. Computational techniques are integrated into the text. Takes a very practical approach, suitable for statistically-minded biologists. Supported by a Website featuring colour images, software, and data sets. Primarily aimed at statistically-minded biologists, bioinformaticians, biostatisticians, and computer scientists working with microarray data, the book is also suitable for postgraduate students of bioinformatics.

Author Biography

Ernst Wit is the author of Statistics for Microarrays: Design, Analysis and Inference, published by Wiley.

John McClure is the author of Statistics for Microarrays: Design, Analysis and Inference, published by Wiley.

Table of Contents

Preface	xi
1 Preliminaries	1	(112)
1.1 Using the R Computing Environment	1	(2)
1.1.1 Installing smida	2	(1)
1.1.2 Loading smida	3	(1)
1.2 Data Sets from Biological Experiments	3	(132)
1.2.1 Arabidopsis experiment: Anna Amtmann	4	(2)
1.2.2 Skin cancer experiment: Nighean Barr	6	(1)
1.2.3 Breast cancer experiment: John Bartlett	7	(2)
1.2.4 Mammary gland experiment: Gusterson group	9	(1)
1.2.5 Tuberculosis experiment: ΒμG@S group	10	(103)
I Getting Good Data	113	(22)
2 Set-up of a Microarray Experiment	15	(8)
2.1 Nucleic Acids: DNA and RNA	15	(1)
2.2 Simple cDNA Spotted Microarray Experiment	16	(7)
2.2.1 Growing experimental material	17	(1)
2.2.2 Obtaining RNA	17	(1)
2.2.3 Adding spiking RNA and poly-T primer	18	(1)
2.2.4 Preparing the enzyme environment	19	(1)
2.2.5 Obtaining labelled cDNA	19	(1)
2.2.6 Preparing cDNA mixture for hybridization	19	(1)
2.2.7 Slide hybridization	20	(3)
3 Statistical Design of Microarrays	23	(34)
3.1 Sources of Variation	24	(2)
3.2 Replication	26	(10)
3.2.1 Biological and technical replication	27	(2)
3.2.2 How many replicates?	29	(1)
3.2.3 Pooling samples	30	(6)
3.3 Design Principles	36	(4)
3.3.1 Blocking, crossing and randomization	37	(2)
3.3.2 Design and normalization	39	(1)
3.4 Single-channel Microarray Design	40	(4)
3.4.1 Design issues	41	(1)
3.4.2 Design layout	42	(1)
3.4.3 Dealing with technical replicates	42	(2)
3.5 Two-channel Microarray Designs	44	(13)
3.5.1 Optimal design of dual-channel arrays	44	(6)
3.5.2 Several practical two-channel designs	50	(7)
4 Normalization	57	(46)
4.1 Image Analysis	57	(5)
4.1.1 Filtering	58	(2)
4.1.2 Gridding	60	(1)
4.1.3 Segmentation	61	(1)
4.1.4 Quantification	62	(1)
4.2 Introduction to Normalization	62	(7)
4.2.1 Scale of gene expression data	63	(2)
4.2.2 Using control spots for normalization	65	(1)
4.2.3 Missing data	65	(4)
4.3 Normalization for Dual-channel Arrays	69	(24)
4.3.1 Order for the normalizations	70	(1)
4.3.2 Spatial correction	71	(5)
4.3.3 Background correction	76	(4)
4.3.4 Dye effect normalization	80	(4)
4.3.5 Normalization within and across conditions	84	(9)
4.4 Normalization of Single-channel Arrays	93	(10)
4.4.1 Affymetrix data structure	93	(1)
4.4.2 Normalization of Affymetrix data	94	(9)
5 Quality Assessment	103	(22)
5.1 Using MIAME in Quality Assessment	104	(1)
5.1.1 Components of MIAME	104	(1)
5.2 Comparing Multivariate Data	105	(8)
5.2.1 Measurement scale	105	(1)
5.2.2 Dissimilarity and distance measures	106	(5)
5.2.3 Representing multivariate data	111	(2)
5.3 Detecting Data Problems	113	(10)
5.3.1 Clerical errors	114	(3)
5.3.2 Normalization problems	117	(2)
5.3.3 Hybridization problems	119	(2)
5.3.4 Array mishandling	121	(2)
5.4 Consequences of Quality Assessment Checks	123	(2)
6 Microarray Myths: Data	125	(12)
6.1 Design	125	(4)
6.1.1 Single-versus dual-channel designs?	125	(4)
6.1.2 Dye-swap experiments	129	(1)
6.2 Normalization	129	(8)
6.2.1 Myth: 'microarray data is Gaussian'	129	(2)
6.2.2 Myth: 'microarray data is not Gaussian'	131	(1)
6.2.3 Confounding spatial and dye effect	132	(1)
6.2.4 Myth: 'non-negative background subtraction'	133	(2)
II Getting Good Answers	135	(116)
7 Microarray Discoveries	137	(40)
7.1 Discovering Sample Classes	137	(18)
7.1.1 Why cluster samples?	138	(1)
7.1.2 Sample dissimilarity measures	139	(5)
7.1.3 Clustering methods for samples	144	(11)
7.2 Exploratory Supervised Learning	155	(5)
7.2.1 Labelled dendrograms	156	(1)
7.2.2 Labelled PAM-type clusterings	157	(3)
7.3 Discovering Gene Clusters	160	(17)
7.3.1 Similarity measures for expression profiles	160	(3)
7.3.2 Gene clustering methods	163	(14)
8 Differential Expression	177	(34)
8.1 Introduction	177	(2)
8.1.1 Classical versus Bayesian hypothesis testing	177	(2)
8.1.2 Multiple testing 'problem'	179	(1)
8.2 Classical Hypothesis Testing	179	(17)
8.2.1 What is a hypothesis test?	180	(3)
8.2.2 Hypothesis tests for two conditions	183	(9)
8.2.3 Decision rules	192	(3)
8.2.4 Results from skin cancer experiment	195	(1)
8.3 Bayesian Hypothesis Testing	196	(15)
8.3.1 A general testing procedure	200	(3)
8.3.2 Bayesian τ-test	203	(8)
9 Predicting Outcomes with Gene Expression Profiles	211	(36)
9.1 Introduction	211	(7)
9.1.1 Probabilistic classification theory	212	(5)
9.1.2 Modelling and predicting continuous variables	217	(1)
9.2 Curse of Dimensionality: Gene Filtering	218	(5)
9.2.1 Use only significantly expressed genes	218	(2)
9.2.2 PCA and gene clustering	220	(2)
9.2.3 Penalized methods	222	(1)
9.2.4 Biological selection	222	(1)
9.3 Predicting Class Memberships	223	(12)
9.3.1 Variance-bias trade-off in prediction	223	(4)
9.3.2 Linear discriminant analysis	227	(4)
9.3.3 kappa-nearest neighbour classification	231	(4)
9.4 Predicting Continuous Responses	235	(12)
9.4.1 Penalized regression: LASSO	235	(8)
9.4.2 kappa-nearest neighbour regression	243	(4)
10 Microarray Myths: Inference	247	(4)
10.1 Differential Expression	247	(2)
10.1.1 Myth: 'Bonferroni is too conservative'	247	(1)
10.1.2 FPR and collective multiple testing	248	(1)
10.1.3 Misinterpreting FDR	248	(1)
10.2 Prediction and Learning	249	(2)
10.2.1 Cross-validation	249	(2)
Bibliography	251	(8)
Index	259