Computational Biology (eBook)

Computational Biology 3
1 Identification of Relevant Genes from Microarray Experiments based on Partial Least Squares Weights: Application to Cancer Genomics 9
1.1 Introduction 9
1.2 Methods 11
1.2.1 Partial Least Squares Dimension Reduction 11
1.2.2 Variable Selection Measures as Functions of PLS Weights 11
1.2.3 Variable Influence Projection in Partial Least Squares 12
1.2.4 B-Partial Least Squares (B-PLS) Regression Coefficient 12
1.2.5 Random Augmentation VIP 13
1.3 Design of Simulation Studies 14
1.3.1 Simulation Based on Normal Model with Cluster-Specific Correlation 15
1.3.2 Resampling-Based Simulation from Real Data 16
1.4 Simulation Results 17
1.4.1 Result Based on Normal Model with Cluster-Specific Correlation 17
1.4.2 Resampling-Based Simulation Result 19
1.5 Applications to Microarray Gene Expression Data 20
1.6 Discussion 21
References 25
2 Geometric Biclustering and Its Applications to Cancer Tissue Classification Based on DNA Microarray Gene Expression Data 26
2.1 Introduction 26
2.2 Geometric Biclustering Patterns 29
2.2.1 Bicluster Types 29
2.2.2 Geometric Expressions of Biclusters 32
2.3 Geometric Biclustering Algorithms 33
2.3.1 Hough Transformation for Line Detection 34
2.3.1.1 The Classical Hough Transformation 34
2.3.1.2 Generalization of the Hough Transformation 35
2.3.2 Geometric Biclustering Algorithm 37
2.3.2.1 Additive and Multiplicative Pattern Plot 37
2.3.2.2 GBC Algorithm 38
2.3.2.3 Applications 40
2.3.3 Relaxation-Based Geometric Biclustering Algorithm 44
2.3.3.1 Nonlinear Probabilistic Relaxation Labeling 44
2.3.3.2 Algorithms 46
2.3.3.3 Applications 48
2.3.4 Geometric Biclustering Using Functional Modules (GBFM) 51
2.3.4.1 Gene Annotation and Functional Modules 51
2.3.4.2 Algorithms 54
2.3.4.3 Applications 55
2.4 Conclusions 58
References 59
3 Statistical Analysis on Microarray Data: Selection of Gene Prognosis Signatures 61
3.1 Introduction 61
3.1.1 Notation 62
3.2 Supervised Classification 62
3.2.1 Linear Classifier 63
3.2.2 Support Vector Machines 63
3.2.3 Nearest Centroid 64
3.2.4 Classification and Regression Trees 65
3.2.5 Error Rate Estimation 66
3.2.5.1 Apparent Error Rate 66
3.2.5.2 Cross-Validation 66
3.2.5.3 Bootstrap Approach 66
3.3 Variable Selection 67
3.3.1 Filter, Wrapper and Embedded Approaches 68
3.3.2 Recursive Feature Elimination 69
3.3.3 Nearest Shrunken Centroids 70
3.3.4 Random Forests 70
3.3.5 Extension to Multiclass 72
3.3.5.1 Division into Binary Problems 72
3.3.5.2 Unbalanced Multiclass Problems 72
3.3.6 Selection Bias and Performance Assessment 73
3.3.7 Optimal Size of the Selection 74
3.4 Illustrative Example with the Golub Data Set 74
3.4.1 Performance of the Three Feature Selection Methods 74
3.4.2 Comparison of the Gene Selections 76
3.4.3 Choice of Method 78
3.5 Validation 78
3.5.1 Biological Interpretation 78
3.5.2 Independent Test Set 79
3.6 Conclusion 79
References 80
4 Agent-Based Modeling of Ductal Carcinoma In Situ: Application to Patient-Specific Breast Cancer Modeling 83
4.1 Introduction 83
4.1.1 Biology of Breast Duct Epithelium 84
4.1.2 Pathobiology of DCIS 86
4.1.3 A Mini-Review of DCIS Modeling 87
4.1.4 Why Agent-Based Modeling? 89
4.2 Agent-Based Model of DCIS 90
4.2.1 A Brief Review of Exponential Random Variables and Poisson Processes 92
4.2.2 A Family of Potential Functions 93
4.2.3 Cell States 94
4.2.3.1 Quiescent Cells (Q) 94
4.2.3.2 Proliferation (P) 95
4.2.3.3 Apoptosis (A) 96
4.2.3.4 Necrosis (N) 97
4.2.3.5 Calcified Debris (C) 98
4.2.4 Cell Motion Based upon the Balance of Forces 98
4.2.4.1 Cell–Cell Adhesion (Fcca) 99
4.2.4.2 Cell–BM Adhesion (Fcba) 99
4.2.4.3 (Calcified) Debris–(Calcified) Debris Adhesion (Fdda) 100
4.2.4.4 Cell–Cell Repulsion (Including Calcified Debris) (Fccr) 100
4.2.4.5 Cell–BM Repulsion (Including Debris) (Fcbr) 100
4.2.5 Duct Geometry 101
4.2.6 Intraductal Oxygen Diffusion 101
4.3 Numerical Technique 102
4.3.1 Efficient Interaction Testing 103
4.4 Estimating Key Parameters 104
4.4.1 Cell Cycle and Apoptosis Time 104
4.4.2 Oxygen Parameters 105
4.4.3 Cell Mechanics 105
4.5 Application to Patient-Specific Modeling 105
4.5.1 Data Sources and Processing 106
4.5.2 Patient-Specific Calibration 106
4.5.3 Verification of Calibration 108
4.5.4 Sample Applications of the Calibrated Model 108
4.5.4.1 Parameter Study: Necrosis and Calcification Time 108
4.6 Ongoing and Future Work 112
References 113
5 Multicluster Class-Based Classification for the Diagnosis of Suspicious Areas in Digital Mammograms 118
5.1 Introduction 118
5.1.1 Background 118
5.1.2 Review of Existing Techniques 120
5.2 Research Methodology 122
5.2.1 Acquiring and Processing of Digital Mammograms 122
5.2.2 Creation of Multicluster Classes with Strong Clusters 123
5.2.3 Classification 124
5.2.3.1 Original Inputs with Multiple Classes 124
5.2.3.2 Cluster Values with Multiple Classes 125
5.3 Experimental Results and Comparative Analysis 125
5.4 Conclusions 127
References 127
6 Analysis of Cancer Data Using Evolutionary Computation 129
6.1 Introduction 129
6.2 Overview of Evolutionary Computation 133
6.2.1 Genetic Programming 133
6.2.1.1 Operations for Modifying the Tree 134
6.2.1.2 Control Parameters 135
6.2.2 Genetic Algorithms 135
6.2.3 Parallel Evolutionary Computation 136
6.2.3.1 Parallelism at Fitness Level 136
6.2.3.2 Parallelism at Population Level (Island Model or Cellular Model) 137
6.2.3.3 Parameters of Island Model 137
6.2.3.4 Application Program Interface Tools 139
6.3 Analysis of Cancer Data 140
6.3.1 Genetic Programming for Binary Classification 140
6.3.1.1 Method 141
6.3.1.2 Experiments 142
6.3.2 Genetic Algorithms for Binary Classification 143
6.3.2.1 Concepts from Geometry 143
6.3.2.2 Nonlinear Programming Problem 144
6.3.2.3 Prediction 144
6.3.2.4 Solving Nonlinear Programming Problem by Genetic Algorithms 144
6.3.2.5 Experiments 145
6.3.3 Genetic Algorithm for Single-Class Classification 146
6.3.3.1 Nonlinear Programming Problem 147
6.3.3.2 Prediction 147
6.3.3.3 Using GA to Solve Nonlinear Programming Problem 147
6.3.3.4 Experiments 148
6.4 Conclusion 149
References 150
7 Analysis of Population-Based Genetic Association Studies Applied to Cancer Susceptibility and Prognosis 152
7.1 Genetic Variation and Its Implication in Cancer 152
7.2 Evolution of Genetic Epidemiology: From Family-Based to Population-Based Association Studies 154
7.3 Technical Issues and Data Quality Control for SNP-Array Association Studies 156
7.3.1 Introduction to Genotype Calling Algorithms 157
7.3.2 SNP-Level Quality Control 159
7.3.2.1 Percentage of Present Calls 159
7.3.2.2 Hardy–Weinberg Equilibrium 159
7.3.2.3 Minor Allele Frequency 160
7.3.2.4 Genotype Calling and Exploration of Signal Intensity Plots 161
7.3.3 Sample-Level Quality Control 162
7.3.3.1 Percentage of Present Calls 162
7.3.3.2 Sample Heterozygosity 162
7.3.3.3 Using Principal Components Analysis as a Method to Detect Outliers or Related Samples 163
7.4 Single-SNP Analysis: Association Between SNPs and a Trait 164
7.4.1 Binary Outcome 165
7.4.2 Quantitative Outcome 166
7.4.3 Prognosis Outcome 167
7.5 Multiple-SNP Analysis 167
7.5.1 Introduction to Haplotypes 168
7.5.2 Linkage Disequilibrium, Linkage Blocks, and Tag-SNPs 168
7.5.3 Haplotype Inference 170
7.5.4 Haplotype Association with Disease 170
7.6 Genome-Wide Association Studies 171
7.6.1 Study Designs 172
7.6.2 Assessing Association in GWAS 176
7.6.3 Statistical Power Calculations 177
7.6.4 Statistical Level Correction for Multiple Testing 178
7.7 Gene–Gene and Gene–Environment Interactions 182
7.8 Bioinformatics Tools and Databases for Genetic Association Studies 183
7.8.1 Genetic Association Suites 183
7.8.1.1 SNPStats 183
7.8.1.2 SNPassoc 184
7.8.1.3 PLINK 184
7.8.1.4 GAP 184
7.8.2 Haplotype-Only Software 185
7.8.2.1 Haploview 185
7.8.2.2 PHASE/fastPHASE 185
7.8.2.3 Haplo.stats 185
7.8.2.4 THESIAS 186
7.8.3 Web Databases 186
7.8.3.1 dbSNP 186
7.8.3.2 Hapmap 187
7.8.3.3 Genome Variation Server 187
7.8.4 Statistical Power Calculation 187
7.8.4.1 QUANTO 187
7.8.4.2 Genetic Power Calculator 188
7.8.4.3 CaTS 188
References 188
8 Selected Applications of Graph-Based Tracking Methods for Cancer Research 195
8.1 Introduction 195
8.2 Object Detection 196
8.3 Local Maximum and Blurring 197
8.4 Object Segmentation 197
8.5 Object Tracking 198
8.6 Algorithms on Graphs 198
8.7 Application to Lamellipodium Dynamics 200
8.8 Application to Mitotic Dynamics 202
8.9 Application to Cell Tracking 203
8.10 Conclusions and Perspectives 204
References 205
9 Recent Advances in Cell Classification for Cancer Research and Drug Discovery 206
9.1 Introduction 206
9.2 Nuclear Segmentation 209
9.2.1 Threshold-Based Segmentation 209
9.2.2 Image Thresholding 210
9.2.3 Fragment Merging Algorithm 210
9.3 Feature Extraction 211
9.3.1 Sequential Forward Selection 211
9.3.2 Automated Feature Weighting 212
9.3.3 Feature Scaling 212
9.4 Cell Phase Modeling 212
9.4.1 Feature Weighting-HMM 214
9.4.2 Feature Weighting-OMM 216
9.4.3 Feature Weighting-GMM 217
9.4.4 Feature Weighting-Fuzzy GMM 217
9.4.5 Feature Weighting-VQ 218
9.4.6 Feature Weighting-Fuzzy VQ 219
9.5 Algorithms for Modeling and Classifying Cell Phases 219
9.5.1 Modeling Algorithm 220
9.5.2 Classification Algorithm 220
9.6 Fuzzy Fusion of Classifiers 221
9.7 Experimental Results 223
9.7.1 Data Set 223
9.7.2 Feature Extraction 223
9.7.3 Initialization and Constraints on Parameters During Training 223
9.7.4 Experimental Results 224
9.8 Conclusion 225
References 225
10 Computational Tools and Resources for Systems Biology Approaches in Cancer 228
10.1 Introduction 228
10.2 Molecular Networks Involved in Cancer 229
10.2.1 Pathways Affected by Cancer Onset and Progression 229
10.2.2 Target Pathways of Cancer Treatment 229
10.3 Molecular Interaction Databases 231
10.3.1 BioCyc 232
10.3.2 KEGG 232
10.3.3 Reactome 232
10.3.4 ConsensusPathDB 233
10.3.5 TRANSPATH® 233
10.3.6 Annotation Tools 233
10.3.7 Modeling Tools 234
10.3.8 Systems Biology Workbench 235
10.3.9 JDesigner 235
10.3.10 CellDesigner 236
10.3.11 PyBioS 236
10.4 Computational Models for Cancer-Related Processes 236
10.4.1 BioModels Database 236
10.4.2 Specific Kinetic Models Relevant for Cancer 237
10.5 Discussion 239
References 240
11 Laser Speckle Imaging for Blood Flow Analysis 244
11.1 Introduction 245
11.2 Experimental Techniques and Setup 246
11.2.1 Experimental Setup 246
11.2.2 Laser Speckle Imaging 247
11.2.2.1 Effect of 2 2 or n n Hardware Binning on the Camera 248
11.2.2.2 Speckle Contrast, K 248
11.2.2.3 Decorrelation Time, c 248
11.2.2.4 Mean Flow Velocity, c 249
11.2.2.5 Parameter, N 250
11.2.3 Laser Speckle Contrast Analysis 250
11.2.4 Spatially Derived Contrast Using Temporal Frame Averaging 251
11.2.5 Temporally Derived Contrast 251
11.2.6 Modified Laser Speckle Imaging 252
11.3 Results and Discussion 252
11.3.1 Effects of Window Size M on KLASCA 253
11.3.2 Effects of n on KsLASCA 255
11.3.3 Effects of n on KtLASCA and NmLSI 255
11.3.4 Comparisons of KLASCA, KsLASCA, and KtLASCA 257
11.3.5 Evaluations on Visual Qualities 259
11.3.5.1 Subjective Quality 259
11.3.5.2 Objective Quality 265
11.3.6 Processing Time 267
11.3.6.1 LASCA 267
11.3.6.2 sLASCA, tLASCA, and mLSI with Varying M and n 267
11.4 Conclusions 269
References 270
12 The Challenges in Blood Proteomic Biomarker Discovery 273
12.1 Introduction 273
12.2 Blood Samples Preparation for Biomarker Discovery 275
12.2.1 Dynamical Range of Proteins 275
12.2.2 The Blood ``Peptidome' 276
12.2.3 Other Biological Factors 277
12.3 Bioinformatics Algorithms in Biomarker Discovery 277
12.3.1 Baseline Removal 278
12.3.2 Denoising/Smoothing 282
12.3.2.1 Discrete Wavelet Transform 282
12.3.2.2 Matched Filtration 282
12.3.2.3 Savitzky–Golay and Average Moving 283
12.3.3 Normalization 283
12.3.3.1 Dividing by a Constant Value 284
12.3.3.2 Regression 284
12.3.3.3 Quantile 285
12.3.4 Peak Detection/Identification 285
12.3.5 Peak Alignment 287
12.3.6 Biomarker Candidate Identification 288
12.3.7 Clinical Diagnosis 289
12.3.8 Protein/Peptide Identification 290
12.4 Validation and Clinical Application 291
References 294
Index 300