Data Science, AI, and Machine Learning in Drug Development

Cover
Half Title
Series Page
Title Page
Copyright Page
Table of Contents
Preface
Contributors
1. Transforming Pharma with Data Science, AI and Machine Learning
1.1 Introduction
1.2 Productivity Challenge
1.3 Shifting Paradigm of Drug Development and Healthcare
1.3.1 Patient-Centricity
1.3.2 Diagnosis versus Prediction
1.3.3 Evidence Generation
1.3.4 Healthcare Ecosystem
1.4 Powering Innovation through Big Data and Digital Technologies
1.4.1 Big Data
1.4.2 Data Strategy
1.4.3 AI and Machine Learning-Enabled Redesign
1.4.3.1 Drug Discovery
1.4.3.2 Clinical Trial
1.4.3.3 Manufacturing
1.4.3.4 Comparative Effectiveness
1.5 Challenges
1.5.1 Data Integration
1.5.2 Infrastructure
1.5.3 Methodological Barriers
1.5.4 Culture and Talent
1.5.5 Regulatory Framework
1.6 Concluding Remarks
References
2. Regulatory Perspective on Big Data, AI, and Machining Learning
2.1 Introduction
2.2 Background
2.2.1 What Is Big Data
2.2.2 Real-World Evidence
2.2.3 AI and Machine Learning
2.2.4 Regulatory Challenges
2.3 FDA Perspective
2.3.1 Critical Path Initiative
2.3.2 The 21st-century Cures Act
2.3.3 Regulating AI/ML-Based Devices
2.3.3.1 Digital Health Innovation Action Plan
2.3.3.2 AI/ML-Based Software as Medical Device Action Plan
2.4 European Medicines Agency Big Data Initiatives
2.4.1 EMA’s View of RWD
2.4.2 EU Medical Device Regulatory Framework
2.5 Other Regulatory Perspectives
2.5.1 Big Data–Based Evidence Generation
2.5.1.1 Health Canada
2.5.1.2 Japan Pharmaceuticals and Medical Devices Agency
2.5.1.3 Other Countries
2.6 AI-Enabled Opportunities in Regulatory Submissions
2.6.1 FDA
2.6.1.1 Applications of AI/ML in FDA Filings
2.6.1.2 Use of AI/ML Tools in FDA’s Review and Research.
2.6.2 EMA
2.6.2.1 Approvals Supported by RWE
2.6.2.2 Approvals of AI/ML-Based Medical Devices in Europe
2.7 Concluding Remarks
References
3. Building an Agile and Scalable Data Science Organization
3.1 Introduction
3.2 History
3.3 Definition of Data Science
3.4 Data Science Versus Statistics
3.4.1 Data
3.4.2 Tools of Data Analysis
3.4.3 Types of Problems
3.4.4 Skills
3.5 Building Data Science Organization
3.5.1 What Is Data-Driven
3.5.2 What Is Needed to Become Data-Driven
3.5.2.1 Data Governance
3.5.2.2 Data Architecture
3.5.2.3 Advanced Analytics
3.5.2.4 People and Culture
3.6 Setting Up a Data Science Organization
3.6.1 Centralized Model
3.6.2 Decentralized Model
3.6.3 Hybrid Model
3.6.4 Metrices of Organization Build
3.7 How to Make It Happen
3.7.1 Vision
3.7.2 Value Proposition
3.7.3 Organizational Model
3.7.4 Strategic Priorities
3.7.4.1 Data Foundation
3.7.4.2 Analytics Build
3.7.5 Talent Development and Acquisition
3.7.6 Change Management
3.8 Concluding Remarks
References
4. AI and Machine Learning in Drug Discovery
4.1 Introduction
4.2 Drug Discovery
4.2.1 Background
4.2.2 Target Identification
4.2.3 Lead Compound Generation
4.2.4 Lead Selection
4.2.5 Preclinical Studies
4.2.6 Rational Approach
4.2.7 Biologic Drug Discovery
4.3 AI, Machine Learning, and Deep Learning
4.3.1 AI in Drug Discovery
4.3.2 Basic Concepts of Machine Learning
4.3.3 Deep Learning
4.3.3.1 Basic Concepts
4.3.3.2 Types of Deep Learning Models
4.4 Applications of AI and Machine Learning in Drug Discovery
4.4.1 Target Identification and Validation
4.4.2 Molecular Machine Learning
4.4.3 Compound Property and Activity Prediction
4.4.4 De Novo Drug Design
4.4.5 Prediction of Drug–Target Interaction
4.4.6 Chemical Synthesis Planning
4.4.7 Imaging Data Analysis
4.5 AI Disruptors
4.6 Case Examples
4.6.1 Endotypes in Atopic Dermatitis Using Real-World Data
4.6.1.1 Background
4.6.1.2 Data
4.6.1.3 PCA and Cluster Analysis
4.6.1.4 Results
4.6.2 Building Quantitative Structure–Activity Relationship
4.6.2.1 Background
4.6.2.2 Data
4.6.2.3 Methods
4.6.2.4 Results
4.6.2.5 Discussion
4.7 Concluding Remarks
References
5. Predicting Anticancer Synergistic Activity through Machine Learning and Natural Language Processing
5.1 Introduction
5.2 Combination Therapy
5.3 Synergy Assessment
5.3.1 Isobologram
5.3.2 Bliss Independence
5.3.3 Loewe Additivity
5.3.4 Dose–Response Curves
5.3.5 Methods of Drug Synergy Assessment
5.3.6 Machine Learning for Combination Drug
5.3.7 Public Data on Compound Synergy
5.4 Case Studies
5.4.1 Deep Learning for Predicting Anticancer Drug Synergy
5.4.1.1 Background
5.4.1.2 Deep Synergy
5.4.1.3 Modeling Strategy
5.4.1.4 Training Data
5.4.1.5 Input Variables
5.4.1.6 Response
5.4.1.7 Deep Learning Model Building and Validation
5.4.1.8 Results
5.4.2 Predicting the Efficacy of Combination Therapy Using Clinical Data
5.4.2.1 Background
5.4.2.2 Clinical Data Extraction
5.4.2.3 Building of Machine Learning Models from Clinical Combination Data
5.4.2.4 Results
5.4.2.5 Discussion
5.5 Concluding Remarks
References
6. AI-Enabled Clinical Trials
6.1 Introduction
6.2 Clinical Development
6.3 Reshaping Clinical Trials
6.4 AI-Enabled Applications in Clinical Trials
6.4.1 Target Product Profile
6.4.2 Clinical Study Design
6.4.2.1 Trial Feasibility
6.4.2.2 Enrichment Design
6.4.2.3 Individualized Treatment
6.4.2.4 Pragmatic Studies
6.4.2.5 External Control Using RWE
6.4.2.6 Prediction of Clinical Trial Success
6.4.3 Study Execution
6.4.3.1 Patient Recruitment, Monitoring, Adherence, and Retention
6.4.3.2 Drug Supply Planning
6.5 Case Study
6.5.1 Patient Selection in Oncology Trials
6.5.1.1 Challenges in Patient Selection
6.5.1.2 Prognostic Scores
6.5.1.3 3i Score
6.5.1.4 Summary
6.6 Concluding Remarks
References
7. Machine Learning for Precision Medicine
7.1 Introduction
7.2 Machine Learning–Based Prognostic Markers
7.2.1 Prognostic Marker Discovery
7.2.1.1 Marker Importance Identification
7.2.1.2 Extension to Time-to-Event Endpoints
7.2.2 Prognostic Marker–Based Subgroup Identification
7.2.2.1 A Framework for Robust Cutoff Derivation
7.3 Machine Learning–Based Predictive Markers
7.3.1 Predictive Marker Discovery
7.3.1.1 Approaches Based on Modified Machine Learning
7.3.1.2 Approaches Based on Causal Inference Machine Learning
7.3.2 Predictive Marker–Based Subgroup Identification
7.3.2.1 Subgroup Identification Algorithms
7.3.2.2 Subgroup (Classifier) Performance Evaluation
7.3.2.3 Case Study: Optimizing a Long-Term Treatment Strategy for Humira HS Patients
7.4 Machine Learning–Based Precision Dosing
7.5 Concluding Remarks
References
8. Reinforcement Learning in Personalized Medicine
8.1 Introduction
8.2 Personalized Medicine Versus Precision Medicine
8.3 Evolution of Personalized Medicine
8.3.1 Subgroup Analysis
8.3.2 Subgroup Identification
8.3.3 Individualized Treatment Recommendation
8.4 Reinforcement Learning in Personalized Medicine
8.4.1 Basic Concepts of Reinforcement Learning
8.4.2 Applications to Clinical Trial and Healthcare
8.5 Reinforcement Learning for Treating Sepsis
8.5.1 Background
8.5.2 Data
8.5.3 State, Action and Reward
8.5.4 Results
8.5.5 Other Reinforcement Learning Methods for Sepsis Treatment
8.6 Discussion
8.7 Concluding Remarks
References
9. Leveraging Machine Learning, Natural Language Processing, and Deep Learning in Drug Safety and Pharmacovigilance
9.1 Introduction
9.2 A Historical Perspective on Drug Safety and Pharmacovigilance.
9.3 Drug Safety Versus Pharmacovigilance
9.4 Importance of Safety Monitoring and Challenges
9.5 Taking a Closer Look at Drug Safety in Product Lifecycle
9.5.1 Drug Safety in Discovery and Preclinical Testing
9.5.2 Safety Assessment in Clinical Development
9.5.2.1 Phase I Clinical Trial
9.5.2.2 Phase II and III Clinical Trial
9.5.2.3 Phase IV Clinical Trial
9.6 Data Sources and Databases for Drug Safety and Pharmacovigilance
9.6.1 Data Sources for Drug Safety and Pharmacovigilance
9.6.2 Databases for Drug Safety and Pharmacovigilance
9.6.2.1 Spontaneous Reporting Systems Safety Databases
9.6.3 Social Media and Web Data
9.6.4 Biomedical Literature
9.6.5 Electronic Health and Medical Records
9.6.6 Clinical Registries
9.7 Drug Safety in the Big Data Paradigm
9.7.1 What Is Big Data?
9.7.2 Big Data in Healthcare and Drug Safety
9.8 New Opportunities in Drug Safety and Pharmacovigilance
9.8.1 Machine Learning
9.8.2 Artificial Intelligence
9.9 Natural Language Processing
9.9.1 Deep Learning
9.9.2 Integration of ML, AI, NLP, and DL in Drug Safety and Pharmacovigilance
9.10 Emerging Regulatory Guidance and Initiatives
9.11 Data Visualization in Drug Safety
9.11.1 Bubble Plot
9.11.2 Chord Diagram
9.12 Some Examples of the Use of Machine Learning, NLP, and AI
9.12.1 Use of Machine Learning, NLP, and AI in Drug-Induced Liver Safety
9.12.2 Other Recent Examples of Machine Learning, NLP, and AI in Drug Safety
9.13 Software and Tools
9.14 Some Limitations of Machine Learning, Artificial Intelligence, Natural Language Processing, and Deep Learning
9.15 Concluding Remarks
References
10. Intelligent Manufacturing and Supply of Biopharmaceuticals
10.1 Introduction
10.2 Biopharmaceutical Manufacturing and Supply
10.2.1 Manufacturing Risks
10.2.2 Supply Chain Challenges
10.3 AI-Powered Opportunities
10.3.1 Intelligent Automation
10.3.2 Multivariate Analytics
10.3.2.1 Process Understanding
10.3.2.2 Process Monitoring and Control
10.3.3 Predictive Maintenance
10.3.4 Demand Forecasting, Logistics, and Inventory Management
10.3.5 End-to-End Visibility
10.4 Case Example
10.4.1 Background
10.4.2 Modeling Approach
10.5 Discussion
10.6 Concluding Remarks
References
11. Reinventing Medical Affairs in the Era of Big Data and Analytics
11.1 Introduction
11.2 Traditional Role of Medical Affairs
11.3 Changing Landscape
11.4 Emerging New Opportunities
11.4.1 Early Discovery Research
11.4.2 Clinical Development
11.4.3 Marketing Application
11.4.4 Product Launch
11.4.5 Lifecycle Management
11.4.6 Stakeholder Engagement
11.5 Case Examples
11.5.1 FASENRA®
11.5.2 ELEN Index
11.5.2.1 Background
11.5.2.2 Method Development
11.5.2.3 Validation
11.5.2.4 Clinical Utility
11.6 Concluding Remarks
References
12. Deep Learning with Electronic Health Record
12.1 Introduction
12.2 Electronic Health Records
12.2.1 A Brief History
12.2.2 EHR, Medical Claim, and Disease Registry
12.2.3 Limitations
12.2.3.1 Privacy and Security
12.2.3.2 Cost
12.2.3.3 Data Quality
12.2.3.4 Interoperability
12.2.3.5 Analysis Challenges
12.3 Applications of EHR
12.3.1 Nature History of Disease
12.3.2 Phenotyping
12.3.3 Risk Prediction and Biomarker Discovery
12.3.4 Evidence-Based Medicine
12.3.5 Detection of Adverse Drug Reaction
12.4 Case Studies
12.4.1 Observational Study of Trauma Care
12.4.1.1 Background
12.4.1.2 Study Design
12.4.1.3 Propensity Score Matching
12.4.1.4 Results
12.4.2 Prediction of Cardiovascular Risk Using EHR Data and Machine Learning
12.4.2.1 Background
12.4.2.2 Data Source
12.4.2.3 Study Population
12.4.2.4 Assessments
12.4.2.5 Machine Learning Algorithms
12.4.2.6 Results
12.5 Concluding Remarks
References
13. Real-World Evidence for Treatment Access and Payment Decisions
13.1 Introduction
13.2 Market Access
13.3 Health Technology Assessment
13.3.1 Price Negotiation
13.3.1.1 Outcome-Based Contracts
13.3.1.2 Risk-Sharing Agreements
13.3.1.3 Alternative/Innovative Payment Models
13.3.2 Frameworks for Value Assessment
13.3.3 Process and Scope of HTA
13.4 RWE for Value Assessment
13.4.1 Comparative Effectiveness
13.4.2 Product Safety
13.4.3 Cost-Effective Analysis
13.4.4 Burden of Disease
13.5 Statistical Methods for Evidence Synthesis
13.5.1 Meta-Analysis
13.5.2 Bayesian Methods
13.5.2.1 Bayes Theorem and Posterior Distribution
13.5.2.2 Inference about Parameters
13.5.2.3 Inference of Future Observations
13.5.3 Survival Analysis
13.5.3.1 Kaplan–Meier Estimation
13.5.3.2 Log-Rank Test
13.5.3.3 Cox’s Proportional Hazards Model
13.6 Case Example
13.6.1 Use of RWE to Extrapolate Long-Term Survival
13.6.1.1 Background
13.6.1.2 Long-Term Survival
13.7 Concluding Remarks
References
Index