Statistical Analysis of Microbiome Data (Frontiers in Probability and the Statistical Sciences)

✍ Scribed by Somnath Datta (editor), Subharup Guha (editor)

• Publisher: Springer
• Year: 2021
• Tongue: English
• Leaves: 349
• Category: Library

✦ Synopsis

Microbiome research has focused on microorganisms that live within the human body and their effects on health. During the last few years, the quantification of microbiome composition in different environments has been facilitated by the advent of high throughput sequencing technologies. The statistical challenges include computational difficulties due to the high volume of data; normalization and quantification of metabolic abundances, relative taxa and bacterial genes; high-dimensionality; multivariate analysis; the inherently compositional nature of the data; and the proper utilization of complementary phylogenetic information. This has resulted in an explosion of statistical approaches aimed at tackling the unique opportunities and challenges presented by microbiome data.

This book provides a comprehensive overview of the state of the art in statistical and informatics technologies for microbiome research. In addition to reviewing demonstrably successful cutting-edge methods, particular emphasis is placed on examples in R that rely on available statistical packages for microbiome data. With its wide-ranging approach, the book benefits not only trained statisticians in academia and industry involved in microbiome research, but also other scientists working in microbiomics and in related fields.

✦ Table of Contents

Preface
Acknowledgments
Contents
Part I Preprocessing and Bioinformatics Pipelines
Denoising Methods for Inferring Microbiome Community Content and Abundance
1 Introduction
2 Common Algorithmic Denoising Strategies
3 Model-Based Denoising
3.1 Hierarchical Divisive Clustering
3.2 Finite Mixture Model
3.3 Denoising Long-Read Technology
4 Model Assessment
4.1 With Known Truth
4.1.1 Accuracy in ASV Identification
4.1.2 Accuracy in Read Assignments
4.2 With Unknown Truth
4.2.1 Assessment with UMIs
4.2.2 Clustering Stability
5 Conclusions
References
Statistical and Computational Methods for Analysis of Shotgun Metagenomics Sequencing Data
1 Introduction
2 Methods for Species Identification and Quantification of Microorganisms
3 Metagenome Assembly and Applications
3.1 de Bruijn Assembly of a Single Genome
3.2 Modification for Metagenome and Metagenome-Assembled Genomes
3.3 Compacted de Bruijn Graph
4 Estimation of Growth Rates for Metagenome-Assembled Genomes (MAGs)
5 Methods for Identifying Biosynthetic Gene Clusters
5.1 A Hidden Markov Model-Based Approach
5.2 A Deep Learning Approach
5.3 BGC Identification Based on Metagenomic Data
6 Future Directions
References
Bioinformatics Pre-Processing of Microbiome Data with An Application to Metagenomic Forensics
1 Introduction
2 Bioinformatics Pipeline
2.1 Microbiome Data
2.2 Quality Control
2.3 Taxonomic Profiling
2.3.1 MetaPhlAn2
2.3.2 Kraken2
2.3.3 Kaiju
2.4 Computing facilities
3 Methodology
3.1 Pre-Processing and Feature Selection
3.2 Exploration of Candidate Classifiers
3.3 The Ensemble Classifier
3.4 Class Imbalance
3.5 Performance Measures
3.6 Data Analysis
4 Results
5 Discussion
6 Data Acknowledgement
7 Code Availability
References
Part II Exploratory Analyses of Microbial Communities
Statistical Methods for Pairwise Comparison of Metagenomic Samples
1 Introduction
2 Microbial Community Comparison Methods Based on OTU Abundance Data
3 Microbial Community Comparison Measures Based on a Phylogenetic Tree
3.1 The Fst Statistic and Phylogenetic Test for Comparing Communities
3.2 UniFrac, W-UniFrac, VAW-UniFrac, and Generalized UniFrac for Comparing Microbial Communities
3.3 VAW-UniFrac for Comparing Communities
4 Alignment-Free Methods for the Comparison of Microbial Communities
5 A Tutorial on the Use of UniFrac Type and Alignment-Free Dissimilarity Measures for the Comparison of Metagenomic Samples
5.1 Analysis Steps for UniFrac, W-UniFrac, Generalized UniFrac, and VAW-UniFrac
5.2 Analysis Steps for the Comparison of Microbial Communities Based on Alignment-Free Methods
6 Discussion
References
Beta Diversity and Distance-Based Analysis of Microbiome Data
1 Introduction
2 Quantifying Dissimilarity: Common Beta Diversity Metrics
3 Ordination and Dimension Reduction
3.1 Principal Coordinates Analysis
3.2 Double Principal Coordinate Analysis
3.3 Biplots
3.4 Accounting for Compositionality
3.5 Model-Based Ordination Using Latent Variables
4 Distance-Based Hypothesis Testing
4.1 Permutation Tests
4.2 Kernel Machine Regression Tests
4.3 Sum of Powered Score Tests
4.4 Adaptive Tests
4.5 Comparison of Distance-Based Tests
5 Strengths, Weaknesses, and Future Directions
References
Part III Statistical Models and Inference
Joint Models for Repeatedly Measured Compositional and Normally Distributed Outcomes
1 Introduction
2 Motivating Data
3 Statistical Models
3.1 The Multinomial Logistic Mixed Model (MLMM)
3.2 Dirichlet-Multinomial Mixed Model (DMMM)
3.3 Goodness of Fit
4 Simulation Studies
4.1 Simulation Setting
4.2 Simulation Results
5 Data Analysis
6 Discussion
7 Software
Appendix
References
Statistical Methods for Feature Identification in Microbiome Studies
1 Introduction
2 Differential Abundance Analysis
2.1 Compositional Methods
2.2 Count-Based Methods
2.3 Additional Notes
3 Mediation Analysis
4 Feature Identification Adjusting for Confounding
4.1 Covariate Adjustment
4.2 Model-Based Standardization
5 Summary
References
Statistical Methods for Analyzing Tree-Structured Microbiome Data
1 Introduction
2 Modeling Multivariate Count Data
2.1 Dirichlet-Multinomial Model
2.2 Dirichlet-Tree Multinomial Model
2.3 Implementation and Illustration
3 Estimating Microbial Compositions
3.1 Empirical Bayes Normalization
3.2 Phylogeny-Aware Normalization
3.3 Statistical Analysis of Compositional Data
3.4 Implementation and Illustration
4 Regression with Compositional Predictors
4.1 Constrained Lasso and Log-Ratio Lasso
4.2 Subcomposition Selection
4.3 Phylogeny-Aware Subcomposition Selection
4.4 Linear Regression and Variable Fusion
5 Additional References
6 Discussion
References
A Log-Linear Model for Inference on Bias in Microbiome Studies
1 Introduction
2 Methods
2.1 The Brooks Data
2.2 Setup and Estimation
2.3 Inference
2.4 Testability of the Hypothesis
2.4.1 Example: Testable Hypotheses for Main Effects
2.4.2 Example: Testable Hypotheses for Interaction Effects
3 Simulations
3.1 Main Effect Simulation
3.2 Interaction Effect Simulation Based on the Brooks Data
4 Results
4.1 Simulation Results
4.2 Do Interactions Between Taxa Affect Bias in the Brooks Data?
4.3 Plate and Sample Type Effects in the Brooks Data
5 Discussion
Appendix
References
Part IV Bayesian Methods
Dirichlet-Multinomial Regression Models with Bayesian Variable Selection for Microbiome Data
1 Introduction
2 Methods
2.1 Dirichlet-Multinomial Regression Models for Compositional Data
2.2 Variable Selection Priors
2.3 Network Priors
2.3.1 Unknown G
2.4 Dirichlet-Tree Multinomial Models
2.5 Posterior Inference
3 Simulated Data
3.1 Simulation Study for DM Regression Models
3.2 DM Sensitivity Analysis
3.3 Simulation Study for DTM Regression Models
3.4 DTM Sensitivity Analysis
4 Applications
4.1 Multi-omics Microbiome Study—Pregnancy Initiative (MOMS-PI)
4.2 Gut Microbiome Study
5 Conclusion
References
A Bayesian Approach to Restoring the Duality Between Principal Components of a Distance Matrix and Operational Taxonomic Units in Microbiome Analyses
1 Introduction
1.1 Motivating Datasets
1.2 Nonlinear or Stochastic Distances
1.3 Limitations of SVD-Based Approaches
2 A Bayesian Formulation
2.1 Posterior Density
3 Model Sum of Squares and Biplots
4 Posterior Inference
4.1 Gibbs Sampler
4.2 Dimension Reduction: Skinny Bayesian Technique
4.2.1 Subsetted Data Matrix
4.2.2 Lower Dimensional Parameters and Induced Posterior
4.2.3 Faster Inference Procedure
4.3 Model Parameter Estimates
5 Simulation Study
5.1 Generation Strategy
6 Data Analysis
6.1 Tobacco Data
6.2 Subway Data
7 Data Acknowledgement
8 Discussion
Supplementary Materials
Appendix
Proof of Lemma 1
References
Part V Special Topics
Tree Variable Selection for Paired Case–Control Studies with Application to Microbiome Data
1 Introduction
2 Gini Index
2.1 Simulation Analysis
3 Multivariate Gini Index
3.1 Conditional Gini Index
4 Variable Importance
5 Analysis of Obesity Using Microbiome Data
6 Discussion
Appendix
References
Networks for Compositional Data
1 Introduction
2 Methods
2.1 Learning Networks from Marginal Associations
2.1.1 ReBoot
2.1.2 SparCC
2.1.3 CCLasso
2.1.4 COAT
2.2 Learning Networks from Conditional Associations
2.2.1 SPIEC-EASI
2.2.2 gCoda
2.2.3 SPRING
3 Data-Generating Models
3.1 Null Models
3.2 Copula Models
3.3 Logistic-Normal Model
4 Results
4.1 Spurious (Partial) Correlations
4.2 Performance in Network Discovery
4.3 Case Studies in R
5 Future Directions
References
Correction to: A Log-Linear Model for Inference on Bias in Microbiome Studies
Index