Statistical Modeling in Biomedical Research: Contemporary Topics and Voices in the Field (Emerging Topics in Statistics and Biostatistics)

Preface
	Part I: Next-Generation Sequence Data Analysis (Chapters 1 –4)
	Part II: Deep Learning, Precision Medicine, and Applications (Chapters 5 –7)
	Part III: Large-Scale Data Analysis and Its Applications (Chapters 8 –11)
	Part IV: Biomedical Research and the Modeling (Chapters 12 –16)
	Part V: Survival Analysis with Complex Data Structure and Its Applications (Chapters 17 –19)
Contents
Contributors
Part I Next Generation Sequence Data Analysis
	Modeling Species Specific Gene Expression Across Multiple Regions in the Brain
		1 Introduction
		2 Methods
			2.1 Overview
			2.2 Use of Information Criterion for Model Selection
			2.3 Estimating Prior Probabilties of Class Membership
			2.4 Empirical Bayes Shrinkage Priors for Negative Binomial Model
			2.5 Leveraging Gene Expression Profiles over Several Brain Regions Using a Markov Random Field
			2.6 Simulations
				2.6.1 Selection of Information Criterion
				2.6.2 Estimation of Markov Random Field Parameters
				2.6.3 Improvement in Performance Due to Markov Random Field
			2.7 Experimental Data
		3 Results
			3.1 Simulation Results
				3.1.1 BIC Produces Best Classifications Overall Under a Variety of Different Scenarios and Parameters
				3.1.2 Estimation of Markov Random Field Parameters is Precise for Exact Priors
				3.1.3 Markov Random Field Can Significantly Improve Classification Errors When Some Neighboring Genes Have Smaller Variance
				3.1.4 Shrunken Priors for RNA-Seq Data Produce More Consistent Results and Reduce the Number of Low-Abundance Genes Found to be Species Specific
				3.1.5 Application to Experimental Data
		4 Discussion
		References
	Classification of EEG Motion Artifact Signals Using Spatial ICA
		1 Introduction
		2 Classification Methods
			2.1 k-Nearest Neighbor (k-nn)
			2.2 Support Vector Machines
			2.3 Naive Bayes
			2.4 Multinomial Logistic Regression
		3 The EEG Motion Artifact Signals Data and Spatial ICA Methodology
		4 Data Analysis Procedure
		5 Classification Results
		6 Discussion
		References
	Weighted K-Means Clustering with Observation Weight for Single-Cell Epigenomic Data
		1 Introduction
		2 Methodology
			2.1 Weighted K-Means
			2.2 Sparse Weighted K-Means
			2.3 Selection of Tuning Parameter for Weighted K-Means
			2.4 Selection of Tuning Parameter for Sparse Weighted K-Means
		3 Simulation
			3.1 Simulation 1: Multivariate Normal Distribution
			3.2 Simulation 2: Dirichlet-Multinomial Distribution
			3.3 Summary
		4 Application to Single-Cell Chromatin Accessibility Data
		5 Conclusion
		Supplementary Information
			Supplementary Notes
			Supplementary Figures (Figs.S1 and S2)
			Supplementary Tables (Tables S1, S2, S3, S4, S5, S6, and S7)
		References
	Discrete Multiple Testing in Detecting Differential Methylation Using Sequencing Data
		1 Introduction
			1.1 Detecting Differential Methylation in Sequencing Data
			1.2 Multiple Testing and False Discovery Rate (FDR)
			1.3 Notations
		2 Conventional FDR Control Procedures
			2.1 The BH Step-up Procedure
			2.2 Storey\'s q-Value Procedure
		3 Control the FDR in Testing Multiple Discrete Hypotheses
			3.1 Modify the Step-Up Sequence
				3.1.1 Gilbert\'s Method
				3.1.2 The BHH Procedure
				3.1.3 The HSU and AHSU Procedure
			3.2 Use Randomized Tests
				3.2.1 Habiger\'s Method
				3.2.2 MCF-Based Procedure
			3.3 Use a Less Upwardly Biased Estimator of π0
				3.3.1 The aBH and aBHH Procedure
				3.3.2 Liang\'s Discrete Right-Boundary Procedure
			3.4 Software Availability
		4 Simulation Study
		5 Conclusion
		References
Part II Deep Learning, Precision Medicine and Applications
	Prediction of Functional Markers of Mass Cytometry Data via Deep Learning
		1 Introduction
		2 Data
			2.1 Pre-processing
			2.2 Exploratory Analysis
		3 Materials and Methods
			3.1 Background on Neural Networks
			3.2 Methods Comparison
		4 Results
		5 Discussion
		References
	Building Health Application Recommender System Using Partially Penalized Regression
		1 Introduction
		2 Partial Regularization via Orthogonality Using the Adaptive Lasso
		3 Simulation Experiments
		4 Real Data Application
		5 Conclusion
		Appendix
		References
	Hierarchical Continuous Time Hidden Markov Model, with Application in Zero-Inflated Accelerometer Data
		1 Introduction
		2 Model Framework
		3 Parameter Estimation
			3.1 Forward-Backward Algorithm
			3.2 Consensus Optimization
			3.3 Theoretical Properties
		4 Simulation Experiments
		5 Application
		6 Conclusions
		References
Part III Large Scale Data Analysis and Its Applications
	Privacy Preserving Feature Selection via Voted Wrapper Method for Horizontally Distributed Medical Data
		1 Introduction
		2 Methods
			2.1 Definition of Privacy
			2.2 Secure Multiparty Computation
			2.3 PAN-SVM Classifier
			2.4 Wrapper Methods
				2.4.1 SVM-RFE
				2.4.2 RSVM
				2.4.3 SVM-t
			2.5 Workflow of PPFSVW
		3 Experiment Results and Discussions
			3.1 Datasets
			3.2 Performance Assessing
			3.3 Feasibility and Effectiveness
			3.4 Comparison with Other Feature Selection Methods
				3.4.1 Classification Accuracy Improvement
				3.4.2 Number of Selected Features
		4 Conclusions
		References
	Improving Maize Trait through Modifying Combination of Genes
		1 Introduction
		2 Materials and Methods
			2.1 Materials
				2.1.1 Data Description
				2.1.2 Data Preprocessing
			2.2 Methods
				2.2.1 Calculation of Bayes Factor
				2.2.2 Permutation Procedure
				2.2.3 Identifying Trait-Dependent Gene Modules and Estimating the Importance of Genes Interacting with Traits
				2.2.4 Extraction of Candidates of Interacting Gene Pairs
				2.2.5 Classification Models
				2.2.6 Significant Estimation of Overlapping Genes and Gene Pairs Across 2 Years
		3 Results
			3.1 Analyze the Trait-Dependent Gene Modules in Maize
			3.2 Functional Enrichment Analysis on Trait-Dependent Gene Modules
			3.3 Validation Results Using Data Sets across 2 Years
			3.4 Candidate Interacting Gene Pairs
			3.5 Prediction Power of the Candidate Interacting Genes
		4 Discussion
		References
	Molecular Basis of Food Classification in TraditionalChinese Medicine
		1 Introduction
		2 Data
		3 Methods
			3.1 Data Imputing
			3.2 Single Factor Analysis
			3.3 Classifier
				3.3.1 SVM
				3.3.2 Ensemble Learning
				3.3.3 Deep Learning
		4 Experiments
			4.1 ANOVA Result
			4.2 TCM Label Classification
		5 Discussion
		References
	Discovery Among Binary Biomarkers in Heterogeneous Populations
		1 Introduction
		2 Logic Regression
		3 Model Description
			3.1 Dirichlet Process Mixture Model (DPMM)
			3.2 Mixture of Finite Mixtures (MFM)
			3.3 The Prior Distribution on the Cluster Assignments
			3.4 Bayesian Joint Model
		4 Estimation
		5 Simulation Study
			5.1 Simulation Results
			5.2 Added Value of the Binary Response
		6 Discussion
		References
Part IV Biomedical Research and the Modelling
	Heat Kernel Smoothing on Manifolds and Its Application to Hyoid Bone Growth Modeling
		1 Introduction
		2 Preliminary
		3 Methods
			3.1 Kernel Smoothing on Manifolds
			3.2 Properties of Kernel Smoothing
			3.3 Numerical Implementation
			3.4 Statistical Inference
			3.5 Validation
		4 Application
			4.1 CT Imaging Data and Preprocessing
			4.2 Results
		5 Conclusions
		References
	Optimal Projections in the Distance-Based Statistical Methods
		1 Introduction
		2 Problem Formulation
		3 Derivable Analytical Results
			3.1 Special Case When the Dimension is 2
			3.2 Second Special Case with Provable Result
		4 Numerical Approach in General Cases
		5 Simulations
			5.1 When the Dimension is 2
			5.2 When We Have n=p
			5.3 General Setting: n > p
		6 Conclusion
		Appendix
			Proof of Theorem 1
			Proof of Theorem 2
			Propositions We Need in Order to Prove Theorem 3
			Proof of Theorem 3
			Proof of Theorem 4
			Proof of Lemma 1
			Proof of Lemma 2
			Proof of Lemma 3
			Proof of Theorem 5
			Proof of Lemma 4
			Proof of Theorem 6
		References
	Kernel Tests for One, Two, and K-Sample Goodness-of-Fit: State of the Art and Implementation Considerations
		1 Introduction
		2 Relevant Statistical Literature
		3 Relevant Machine Learning Literature: Maximum Mean Discrepancy
		4 Statistical Distance-Based Goodness-of-Fit Tests
			4.1 One-Sample Goodness-of-Fit Tests
			4.2 Two and K-Sample Goodness-of-Fit Tests
		5 Practical Implementation
		6 Simulation Study
			6.1 Simulation Design
			6.2 Simulation Results
				6.2.1 Parametric Case
				6.2.2 Nonparametric Case
		7 Real Data Illustration
		8 Discussion and Conclusions
		References
	Hierarchical Modeling of the Effect of Pre-exposure Prophylaxis on HIV in the US
		1 Introduction
		2 Methods
			2.1 Mathematical Model
			2.2 Preventative Medication Protocol
			2.3 Parameter Estimates
				2.3.1 National Population Dynamics
				2.3.2 Urban Population Dynamics
		3 Results
			3.1 Effect of Preventative Medication
				3.1.1 Model Sensitivity Changes with Preventative Medication
		4 Conclusions
		Appendix
		References
	Mathematical Model of Mouse Ventricular Myocytes Overexpressing Adenylyl Cyclase Type 5
		1 Introduction
		2 Model Development
		3 Method of Simulation
		4 Results
			4.1 cAMP and PKA Activation in WT and TG Myocytes Overexpressing AC5
			4.2 The Effects of Isoproterenol on the Action Potential, Ca2+ and Na+ Dynamics in WT and TG Mouse Ventricular Myocytes Overexpressing AC5
			4.3 Frequency Dependence of the Action Potential and [Ca2+]i Transients in WT and TG Mouse Ventricular Myocytes Overexpressing AC5
			4.4 Simulation of DADs in WT and TG Mouse Ventricular Myocytes Overexpressing AC5
		5 Discussion
		6 Conclusions
		References
Part V Survival Analysis with Complex Data Structure and Its Applications
	Non-parametric Maximum Likelihood Estimation for Case-Cohort and Nested Case-Control Designs with Competing Risks Data
		1 Introduction
		2 Likelihood Function, Score Function and NPMLE
			2.1 Likelihood Function
			2.2 Score Functions
			2.3 The NPMLE
			2.4 The Predicted Cumulative Incidence Function
		3 Large Sample Theory
			3.1 Identifiability
			3.2 Consistency of NPMLE
			3.3 Asymptotic Normality of NPMLE
			3.4 A Profile Likelihood Theory for the Estimation of β
				3.4.1 A Consistent Estimator of the Asymptotic Variance for β
		4 Simulation Study
			4.1 Multiple Outcomes with Time Matching
			4.2 Multiple Outcomes with Stratified Matching
			4.3 Results
		5 Example: Application to Liver Cancer in Type 2 DM Patients
		6 Discussion
		Appendix 1: Proof of Lemma 1
		Appendix 2: Proof of Lemma 6
		Appendix 3: Σ is Positive Definite and Symmetric
		Appendix 4: M (β\"0362βM-β0 ) Has Asymptotic Variance Σ-1
		Appendix 5: Two Results Due to V&W van1996
		Appendix 6: The Conditional Distribution of Z Given Y
		References
	Variable Selection in Partially Linear Proportional Hazards Model with Grouped Covariates and a Diverging Number of Parameters
		1 Introduction
		2 Group Variable Selection in the PL-PHM
		3 Asymptotic Theory
			3.1 A General Theorem
			3.2 Adaptive Hierarchical Penalty
		4 Numerical Results
			4.1 Simulation Studies
			4.2 Application to a Real Data Set
		5 Concluding Remarks
		Appendix
		References
	Inference of Transition Probabilities in Multi-State Models Using Adaptive Inverse Probability Censoring Weighting Technique
		1 Introduction
		2 AIPCW in Multi-State Models
			2.1 Estimate Transition Probabilities For a Three-Level Model
			2.2 Estimate Transition Probabilities For a Four-Level Model
		3 Simulation Studies and Results
			3.1 Compare SIPCW and AIPCW
			3.2 Evaluate the Performance of AIPCW
			3.3 Evaluate the Performance of AIPCW in a Four-Level Twelve-State Model
		4 Real Data Example: The HSCT Data
			4.1 Quantify Transition Probabilities
		5 Discussion
		Appendix 1
		Appendix 2
		References
Index