The study of biological data is constantly undergoing profound changes. Firstly, the volume of data available has increased considerably due to new high throughput techniques used for experiments. Secondly, the remarkable progress in both computational and statistical analysis methods and infrastructures has made it possible to process these voluminous data.
The resulting challenge concerns our ability to integrate these data, i.e. to use their complementary nature effectively in the hope of advancing our knowledge. Therefore, a major challenge in studying biology today is integrating data for the most exhaustive analysis possible.
Biological Data Integration deals in a pedagogical way with research work in biological data science, examining both computational approaches to data integration and statistical approaches to the integration of omics data.
Table of Contents
Preface xi
Christine FROIDEVAUX, Marie-Laure MARTIN-MAGNIETTE and Guillem RIGAILL
Part 1 Knowledge Integration 1
Chapter 1 Clinical Data Warehouses 3
Maxime WACK and Bastien RANCE
1.1 Introduction to clinical information systems and biomedical warehousing: data warehouses for what purposes? 3
1.1.1 Warehouse history 4
1.1.2 Using data warehouses today 4
1.2 Challenge: widely scattered data 5
1.3 Data warehouses and clinical data 6
1.3.1 Warehouse structures 6
1.3.2 Warehouse construction and supply 11
1.3.3 Uses 11
1.4 Warehouses and omics data: challenges 15
1.4.1 Challenges of data volumetry and structuring omic data 16
1.4.2 Attempted solutions 17
1.5 Challenges and prospects 18
1.5.1 Toward general-purpose warehouses 18
1.5.2 Ethical dimension of the implementation and the use of warehouses 19
1.5.3 Origin and reproducibility 19
1.5.4 Data quality 20
1.5.5 Data warehousing federation and data sharing 21
1.6 References 21
Chapter 2 Semantic Web Methods for Data Integration in Life Sciences 25
Olivier DAMERON
2.1 Data-related requirements in life sciences 26
2.1.1 Databases for the life sciences 26
2.1.2 Requirements 27
2.1.3 Common approaches: InterMine and BioMart 30
2.2 Semantic Web 31
2.2.1 Techniques 32
2.2.2 Implementation 42
2.3 Perspectives 43
2.3.1 Facilitating appropriation to users 43
2.3.2 Facilitating the appropriation by software programs: FAIR data 44
2.3.3 Federated queries 45
2.4 Conclusion 46
2.5 References 47
Chapter 3 Workflows for Bioinformatics Data Integration 53
Sarah COHEN-BOULAKIA and Frédéric LEMOINE
3.1 Introduction 53
3.2 Bioinformatics data processing chains: difficulties 54
3.2.1 Designing a data processing chain 55
3.2.2 Analysis execution and reproducibility 56
3.2.3 Maintenance, sharing and reuse 58
3.3 Solutions provided by scientific workflow systems 59
3.3.1 Fundamentals of workflow systems 59
3.3.2 Workflow systems 64
3.4 Use case: RNA-seq data analysis 69
3.4.1 Study description 69
3.4.2 From data processing chain to workflows 72
3.4.3 Data processing chains implemented as workflows: conclusion 75
3.5 Challenges, open problems and research opportunities 77
3.5.1 Formalizing workflow development 77
3.5.2 Workflow testing 78
3.5.3 Discovering and sharing workflows 79
3.6 Conclusion 80
3.7 References 81
Part 2 Integration and Statistics 87
Chapter 4 Variable Selection in the General Linear Model: Application to Multiomic Approaches for the Study of Seed Quality 89
Céline LÉVY-LEDUC, Marie PERROT-DOCKÈS, Gwendal CUEFF and Loïc RAJJOU
4.1 Introduction 90
4.2 Methodology 93
4.2.1 Estimation of the covariance matrix Σq 93
4.2.2 Estimation of B 96
4.3 Numerical experiments 99
4.3.1 Statistical performance 99
4.3.2 Numerical performance 100
4.4 Application to the study of seed quality 103
4.4.1 Metabolomics data 104
4.4.2 Proteomics data 105
4.5 Conclusion 108
4.6 Appendices 108
4.6.1 Example of using the package MultiVarSel for metabolomic data analysis 108
4.6.2 Example of using the package MultiVarSel for proteomic data analysis 110
4.7 Acknowledgments 113
4.8 References 113
Chapter 5 Structured Compression of Genetic Information and Genome-Wide Association Study by Additive Models 117
Florent GUINOT, Marie SZAFRANSKI and Christophe AMBROISE
5.1 Genome-wide association studies 118
5.1.1 Introduction to genetic mapping and linkage analysis 118
5.1.2 Principles of genome-wide association studies 119
5.1.3 Single nucleotide polymorphism 120
5.1.4 Disease penetrance and odds ratio 122
5.1.5 Single marker analysis 124
5.1.6 Multi-marker analysis 126
5.2 Structured compression and association study 132
5.2.1 Context 132
5.2.2 New structured compression approach 133
5.3 Application to ankylosing spondylitis (AS) 142
5.3.1 Data 142
5.3.2 Predictive power evaluation 143
5.3.3 Manhattan diagram 144
5.3.4 Estimation for the most significant SNP aggregates 144
5.4 Conclusion 146
5.5 References 146
Chapter 6 Kernels for Omics 151
Jérôme MARIETTE and Nathalie VIALANEIX
6.1 Introduction 152
6.2 Relational data 153
6.2.1 Data described by the kernel 153
6.2.2 Data described by a general (dis)similarity measure 155
6.3 Exploratory analysis for relational data 158
6.3.1 Kernel clustering 158
6.3.2 Kernel principal component analysis 161
6.3.3 Kernel self-organizing maps 163
6.3.4 Limitations of relational methods 166
6.4 Combining relational data 168
6.4.1 Data integration in systems biology 168
6.4.2 Kernel approaches in data integration 169
6.4.3 A consensual kernel 172
6.4.4 A parsimonious kernel that preserves the topology of the initial data 173
6.4.5 A complete kernel preserving the topology of the initial data 175
6.5 Application 176
6.5.1 Loading Tara Ocean data 176
6.5.2 Data integration by kernel approaches 177
6.5.3 Exploratory analysis: kernel PCA 179
6.6 Session information for the results of the example 186
6.7 References 188
Chapter 7 Multivariate Models for Data Integration and Biomarker Selection in ‘Omics Data 195
Sébastien DÉJEAN and Kim-Anh LÊ CAO
7.1 Introduction 195
7.2 Background 197
7.2.1 Mathematical notations 197
7.2.2 Terminology 198
7.2.3 Multivariate projection-based approaches 198
7.2.4 A criterion to maximize specific to each methodology 199
7.2.5 A linear combination of variables to reduce the dimension of the data 199
7.2.6 Identifying a subset of relevant molecular features 200
7.2.7 Summary 200
7.3 From the biological question to the statistical analysis 201
7.3.1 Exploration of one dataset: PCA 201
7.3.2 Classify samples: projection to latent structure discriminant analysis 206
7.3.3 Integration of two datasets: projection to latent structure and related methods 210
7.3.4 Integration of several datasets: multi-block approaches 215
7.4 Graphical outputs 220
7.4.1 Individual plots 220
7.4.2 Variable plots 221
7.5 Overall summary 222
7.6 Liver toxicity study 223
7.6.1 The datasets 223
7.6.2 Biological questions and statistical methods 223
7.6.3 Single dataset analysis 224
7.6.4 Integrative analysis 231
7.7 Conclusion 238
7.8 Acknowledgments 238
7.9 Appendix: reproducible R code 239
7.9.1 Toy examples 239
7.9.2 Liver toxicity 243
7.10 References 247
List of Authors 251
Index 255
