Enables researchers and engineers to gain insights into the capabilities of machine learning approaches to power applications in their fields
Machine Learning and Big Data-enabled Biotechnology discusses how machine learning and big data can be used in biotechnology for a wide breadth of topics, providing tools essential to support efforts in process control, reactor performance evaluation, and research target identification.
Topics explored in Machine Learning and Big Data-enabled Biotechnology include:
- Deep learning approaches for synthetic biology part design and automated approaches for GSM development from DNA sequences
- De novo protein structure and design tools, pathway discovery and retrobiosynthesis, enzyme functional classifications, and proteomics machine learning approaches
- Metabolomics big data approaches, metabolic production, strain engineering, flux design, and use of generative AI and natural language processing for cell models
- Automated function and learning in biofoundries and strain designs
- Machine learning predictions of phenotype and bioreactor performance
Machine Learning and Big Data-enabled Biotechnology earns a well-deserved spot on the bookshelves of reaction, process, catalytic, and environmental engineers seeking to explore the vast opportunities presented by rapidly developing technologies.
Table of Contents:
Preface xv
1 From Genome to Actionable Insights in Biotechnology 1
James Morrissey, Benjamin Strain, and Cleo Kontoravdi
1.1 Introduction 1
1.2 From Genome to Network 2
1.2.1 Metabolic Networks 3
1.2.1.1 Bottom-Up Approaches for Network Reconstruction 3
1.2.1.2 Top-Down Approaches for Network Reconstruction 4
1.2.2 Networks Beyond Metabolism 5
1.3 From Draft to Functional Network 6
1.3.1 Additional Reactions 6
1.3.1.1 Exchange Reactions 6
1.3.1.2 Demand Reactions 6
1.3.1.3 Transport Reactions 6
1.3.1.4 Spontaneous Reactions 7
1.3.1.5 Nongrowth Associated ATP Maintenance 7
1.3.1.6 Biomass Reaction 7
1.3.2 Network Validation 7
1.3.2.1 Manual Screening 8
1.3.2.2 Screening for Dead-End Reactions and Blocked Metabolites 8
1.3.2.3 Infinite Loops 9
1.3.2.4 Leaks and Siphons 10
1.4 From Functional Network to Model 10
1.4.1 Flux Balance Analysis 11
1.4.2 Flux Variability Analysis 12
1.4.3 Flux Sampling 13
1.5 From Model to In Silico Predictions 15
1.5.1 Constraints 15
1.5.2 Objective Function 16
1.5.3 Validating In Silico Predictions 16
1.5.3.1 Growth Rate Predictions 22
1.5.3.2 Amino Acid Auxotrophies 22
1.5.3.3 Gene Essentialities 22
1.5.3.4 Known Host Traits 23
1.5.3.5 Intracellular Predictive Accuracy 23
1.5.4 Toward Multilayer, Multiscale Metabolic Networks 23
1.5.4.1 Integrating Gene Regulatory Networks 24
1.5.4.2 Integrating Transcription and Translation 25
1.5.4.3 Integrating Signaling Networks 25
1.5.4.4 Multicellular and Multitissue Models 25
1.5.4.5 Multiscale Bioreactor Models 26
1.6 From Predictions to Actionable Insights in Biotechnology 26
1.6.1 Metabolic Engineering 26
1.6.2 Cell Line Development and Metabolic Profiling 27
1.6.3 Media and Feed Design 29
1.6.4 Gene Essentiality 30
1.6.5 Kinetic Parameter Estimation 30
1.6.6 Process Monitoring and Forecasting 30
References 31
2 Automated Approaches for the Development of Genome-Scale Metabolic Network Models 43
Emma M. Glass, Deborah A. Powers, and Jason A. Papin
2.1 Introduction 43
2.2 Manual GSM Creation 44
2.3 Automated GSM Development 45
2.3.1 General Approach for Automated GSM Methods 45
2.3.2 GSM Construction Tools 46
2.3.2.1 From Raw Sequences 46
2.3.2.2 From Pre-annotated Sequences 52
2.3.2.3 From Reaction Database Information 53
2.3.2.4 Based on Existing GSMs 56
2.3.2.5 GSM Modification and Visualization Tools 57
2.4 Applications of Automatically-Generated GSM Collections 59
2.4.1 AGORA1 – 773 GSMs 59
2.4.2 EMBL GEMs – 5,587 GSMs 60
2.4.3 MetaGEM – 447 GSMs 61
2.4.4 AGORA2–7,302 GSMs 61
2.4.5 PATHGENN – 914 GSMs 62
2.5 Future Directions for the Field of Automated GSM Development 62
2.6 Conclusion 63
References 63
3 Machine-Guided Approaches for Synthetic Biology Part Design 67
Marc Amil, Leandro N. Ventimiglia, and Aleksej Zelezniak
3.1 Introduction 67
3.2 Model-Guided Sequence Design Using Deep Learning 70
3.2.1 Predictive Models for DNA Function: CNNs in Regulatory Sequence Analysis 70
3.2.1.1 Data Considerations for Supervised Learning on Genomic Sequences 72
3.2.1.2 Primer on Convolutional Neural Networks for Supervised Genomic Sequence Modeling 73
3.2.2 Generative Sequence Modeling 75
3.2.2.1 Data Preparation for Unsupervised Learning 76
3.2.2.2 GANs for the Design of Biological Sequences 77
3.2.2.3 Transformer-Based DNA Models 80
3.2.2.4 Diffusion Models for the Design of Biological Sequences 84
3.3 Sources of Sequence–Function Data for Deep Learning 85
3.3.1 Native-Context Genome-Derived Datasets 86
3.3.2 Synthetic Datasets 87
3.4 Evaluating Synthetic Biological Parts Using Motif Analysis and Deep Learning 90
3.5 Current Challenges of Generative Part Design 93
3.6 Conclusion 93
References 94
4 Machine Learning for Sequence-to-Function Approaches 103
Rana A. Barghout, Maxim Kirby, Austin Zheng, Lya Chinas, Marjan Mohammadi, Zhiqing Xu, Benjamin Sanchez-Lengeling, and Radhakrishnan Mahadevan
4.1 Introduction 103
4.2 Current State of Sequence-to-Function Modeling 105
4.2.1 Protein Function Prediction: From BLAST to Language Models 105
4.2.2 Gene Ontology 106
4.2.3 Enzyme Commission Numbers 106
4.2.4 Enzyme Activity 109
4.2.5 Protein Thermal Stability 109
4.2.6 Protein Toxicity 110
4.2.7 Protein Solubility 111
4.3 Tool Kits and Benchmarks 111
4.3.1 Overview of Open-Source Tools 111
4.3.2 Importance of Standardized Benchmarks 114
4.4 Emerging ML Methods 115
4.4.1 Contrastive Learning 115
4.4.2 Meta Learning 116
4.5 Case Studies 116
4.5.1 Prediction of Enzyme Activity and Substrate Specificity 116
4.5.1.1 Predicting k cat Using CPI-Pred 117
4.6 Challenges in Sequence-to-Function Mapping 118
4.6.1 Sparse Experimental Data 119
4.6.2 Interpretability 120
4.7 Conclusion and Future Directions 120
References 121
5 Prediction of Enzyme Functions by Artificial Intelligence 131
Ha Rim Kim, Hongkeun Ji, Gi Bae Kim, and Sang Yup Lee
5.1 Introduction 131
5.2 Conventional Computational Approaches for Predicting Enzyme Function 132
5.3 Prediction of Enzyme Functions Using Machine Learning 133
5.3.1 Extraction of Enzyme Features from Amino Acid Sequences 134
5.3.2 Machine Learning-Based Approaches Algorithms for Enzyme Function Prediction 136
5.4 Prediction of Enzyme Functions Using Deep Learning 138
5.4.1 Convolutional Neural Network 139
5.4.2 Recurrent Neural Network 141
5.4.3 Transformer and Protein Language Models 142
5.4.4 Graph Neural Network 145
5.5 Concluding Remarks 147
Acknowledgments 153
References 153
6 Design of Biochemical Pathways via AI/ML-Enabled Retrobiosynthesis 161
Hongxiang Li, Xuan Liu, and Huimin Zhao
6.1 Introduction 161
6.1.1 Computer-Aided Synthesis Planning 161
6.1.2 Retrobiosynthesis 162
6.2 Retrobiosynthesis Tools 162
6.2.1 Template-Based Tools 162
6.2.2 Template-Free Tools 166
6.2.3 Searching Algorithm 167
6.2.4 Ranking 168
6.3 Enzyme Selection and Optimization 168
6.3.1 Enzyme Substrate Specificity 169
6.3.2 Enzyme Catalytic Efficiency 170
6.3.3 Enzyme Engineering 172
6.3.4 De Novo Enzyme Design and Discovery 172
6.4 Perspectives 174
6.4.1 Integrating Biocatalysis with Chemocatalysis 175
6.4.2 Toward Next-Generation AI for Retrobiosynthesis Planning 175
6.4.3 Enhancing Enzyme Prediction and Design Capabilities 176
6.4.4 Data Standardization and Model Interpretability 177
Acknowledgments 178
References 179
7 Machine Learning to Accelerate the Discovery of Therapeutic Peptides 183
Nicole Soto-Garcia, Mehdi D. Davari, and David Medina-Ortiz
7.1 Introduction 183
7.2 Peptides: Definitions and Main Characteristics 184
7.3 Benefits and Limitations of Therapeutic Peptides 185
7.4 Computational Design of Therapeutic Peptides 186
7.5 Data Sources for Peptide Discovery 187
7.6 ML-Based Strategies to Accelerate the Discovery of Therapeutic Peptides 189
7.6.1 Data-Driven Approaches 190
7.6.2 ml Strategies for Peptide Bioactivity Classification 192
7.6.2.1 Classification Models for Antimicrobial Peptides 192
7.6.2.2 Classification Models for Antiviral Peptides 193
7.6.2.3 Classification Models for Antifungal Peptides 194
7.6.2.4 Additional Classification Models for Therapeutic Peptides 194
7.6.3 Strategies for Building Toxicity Classification Models 195
7.6.3.1 Toxicity Prediction 196
7.6.3.2 Immunogenicity Identification 197
7.6.3.3 Hemolysis Evaluation 197
7.6.3.4 Other Toxic Adverse Effect Predictions 198
7.6.4 Data-Driven Strategies for Modeling Pharmacological Profiles 198
7.6.5 De novo Design of Therapeutic Peptides 199
7.6.5.1 Variational Autoencoder-Based Approaches 200
7.6.5.2 Generative Adversarial Networks 200
7.6.5.3 Transformer-Based Language Models 201
7.6.5.4 Diffusion Models 201
7.7 Next-Generation Peptide Design Through Multi-Agent Systems 201
7.8 Developing AI-Agent for Autonomous Therapeutic Peptide Design 203
7.9 Conclusion and Perspectives 205
Acknowledgments 206
References 207
8 Machine Learning Approaches for High-Throughput Microbial Identification/Culturing 219
Mohamed Mastouri and Yang Zhang
8.1 Introduction 219
8.2 High-Throughput (HTP) Techniques in Microbial Research 221
8.2.1 Definition and Scope of HTP Microbial Techniques 221
8.2.2 Metagenomics and Next-Generation Sequencing (NGS) 223
8.2.3 Mass Spectrometry-Based Proteomics (MALDI-TOF MS) 223
8.2.4 Flow Cytometry 224
8.2.5 Microfluidics and Lab-on-a-Chip Systems 224
8.2.6 High-Content Imaging and Phenotyping 225
8.3 Fundamentals of Machine Learning 226
8.3.1 Definition of Machine Learning and AI 226
8.3.2 Supervised vs. Unsupervised vs. Reinforcement Learning 226
8.3.3 ML Algorithms Commonly Used in Microbial Identification 229
8.4 Machine Learning Approaches for High-Throughput Microbial Identification 230
8.4.1 Genomic and Metagenomic Data Processing 230
8.4.2 Mass Spectrometry-Based Identification 234
8.4.3 Imaging-Based Identification 236
8.5 Machine Learning Approaches for High-Throughput Microbial Culturing 237
8.5.1 ML-Driven Microbial Growth Prediction 237
8.5.2 AI in Microbial Cultivation Process Optimization 238
8.5.3 Synthetic Biology and AI-Driven Strain Engineering 240
8.6 Challenges and Limitations of Machine Learning in HTP Microbial Research 241
8.7 Future Perspectives and Emerging Trends 242
8.8 Conclusion 243
Acknowledgments 244
References 244
9 Generative AI for Knowledge Mining of Synthetic Biology and Bioprocess Engineering Literature 253
Zhengyang Xiao and Yinjie J. Tang
9.1 Introduction 253
9.2 Text Mining Using Knowledge Graph Tools 254
9.2.1 NEKO: A Lightweight Knowledge Graph Tool 254
9.2.2 GraphRAG 256
9.3 LLM-Automated Data Extraction for Machine Learning 258
9.4 Current Limitations 259
9.5 Conclusion 260
Acknowledgments 260
References 260
10 Metabolomics: Big Data Approaches 263
Kenya Tanaka, Christopher J. Vavricka, and Tomohisa Hasunuma
10.1 Introduction 263
10.2 Methods for Metabolomics 264
10.2.1 Preparation of Samples 264
10.2.2 Detection and Quantification 266
10.3 Analysis and Application of Metabolomics Data for Biotechnology 267
10.3.1 Metabolomics for Identification of Pathway Bottlenecks 267
10.3.2 Absolute Metabolomics and Thermodynamic Analyses 272
10.3.3 Evaluation and Optimization of Metabolic Flux 272
10.4 Artificial Intelligence (AI)-Based Metabolomics Data Processing and Analysis 273
10.5 Future Direction of Metabolomics and Its Analysis 274
References 278
11 Strain Engineering, Flux Design, and Metabolic Production Using Big Data: Ongoing Advances and Opportunities 285
Rafael S. Costa and Rui Henriques
11.1 Introduction 285
11.2 Big Data in Biotechnology: Prerequisites for ML-Based Approaches 287
11.2.1 Data Multimodality and Heterogeneity 287
11.2.2 Stratification and Data Transformations 288
11.2.3 Handling of Missing Values and Outliers 289
11.2.4 Longitudinal Studies 290
11.3 Types of ML-Based Approaches for the Design of Microbial Cell Factories 291
11.3.1 Machine Learning (ML) Models 291
11.3.2 Supervised ml 292
11.3.2.1 Neural Networks 292
11.3.2.2 Decision Trees and Ensembles 293
11.3.2.3 Alternative Predictive Approaches 294
11.3.2.4 Selected Case Studies 294
11.3.3 Unsupervised ml 296
11.3.3.1 Clustering 296
11.3.3.2 Biclustering 297
11.3.3.3 Representation Learning and Dimensionality Reduction 297
11.3.4 Hybrid ML and Constraint-Based Models 298
11.3.4.1 CBM-FBA as Input for ml 299
11.3.4.2 ml as Input for CBM-FBA 299
11.4 ML-Based Approaches in Microbial Cell Factory Design: Case Studies 300
11.4.1 ML-Based Approaches in Strain Engineering and Flux Design 301
11.4.2 Application of ML-Based Approaches for Metabolic Production 304
11.5 Conclusions and Perspectives 306
References 308
12 Next-Generation Metabolic Flux Analysis Using Machine Learning 317
Ahmed Almunaifi, Richard C. Law, Samantha O’Keeffe, Kartikeya Pande, Tongjun Xiang, Onyedika Ukwueze, Aranaa Odai-Okley, Pin-Kuang Lai, and Junyoung O. Park
12.1 Introduction 317
12.2 Dynamic Nature of Metabolism 317
12.3 Flux Balance Analysis and Metabolic Flux Analysis 321
12.4 Incorporating Machine Learning into Metabolic Flux Analysis 324
12.4.1 Challenges in Applying ML to MFA 325
12.4.1.1 A Lack of Isotope Labeling Patterns and Flux Data for Training Machine Learning Models 325
12.4.1.2 Variable-Size Input of Isotope Labeling Patterns for Flux Prediction 326
12.4.1.3 Incorporation of Disparate Data Types 326
12.4.1.4 Initial Computational Cost 326
12.4.2 Available Computational Tools 326
12.4.2.1 Sampling Metabolic Fluxes for Training ML Models 327
12.4.2.2 Atom Mapping Throughout Metabolic Networks 327
12.4.2.3 Selection of Information-Rich Isotope Tracers 327
12.4.2.4 Other Helpful ML Tools 328
12.4.3 A General Workflow for Machine Learning-Based Metabolic Flux Analysis 328
12.4.3.1 Metabolic Model Construction 329
12.4.3.2 Acquiring Training Data 330
12.4.3.3 Training Machine Learning Models 330
12.4.3.4 Evaluation of Machine Learning Models 330
12.4.3.5 Execution: Computing Metabolic Fluxes Directly from Isotope Labeling Patterns 331
12.4.4 Improved Speed and Accuracy of ML-Based MFA 331
12.4.5 Toward Dynamic Flux and Isotope Labeling Analysis 332
12.5 Future Outlook 333
References 334
13 Streamlining the Design-Build-Test-Learn Process in Automated Biofoundries 341
Enrico Orsi, Nicolás Gurdo, and Pablo I. Nikel
13.1 Introduction 341
13.2 The Design-Build-Test-Learn Cycle (DBTLc) 342
13.2.1 The DBTLc Components 342
13.2.2 Application of the DBTLc for Strain Engineering 345
13.2.3 Description of Biofoundries and Their Operating Parts 345
13.2.4 Laboratory Workflows with Potential of Automation in Biofoundries 347
13.3 Geographical Distribution of Biofoundries Around the Globe 350
13.4 Challenges for Implementing Fully Automated Biofoundries 352
13.5 Perspectives and Outlook 355
Acknowledgments 357
Ethics Declarations 357
References 357
14 Machine Learning-Enhanced Hybrid Modeling for Phenotype Prediction and Bioreactor Optimization 367
Oliver Pennington, Yirong Chen, Youping Xie, and Dongda Zhang
14.1 Bioprocess Modeling and Optimization 367
14.1.1 Challenges in Bioprocess Modeling 367
14.1.2 Machine Learning for Bioprocess Modeling 369
14.1.2.1 Principal Component Analysis and Partial Least Squares 369
14.1.2.2 Artificial Neural Networks 370
14.1.2.3 Gaussian Processes 370
14.1.2.4 Ensemble Learning 371
14.1.2.5 Reinforcement Learning 371
14.1.2.6 Future Prospects for Machine Learning 371
14.1.3 Introduction to Hybrid Modeling 372
14.1.3.1 Literature Review 372
14.1.3.2 Hybrid Modeling for Biosystem Optimization 373
14.2 Methodology for Hybrid Modeling 378
14.2.1 Mechanistic Model Construction 378
14.2.2 Time-Varying Parameter Estimation 380
14.2.3 Machine Learning Model Construction 381
14.2.4 Hybrid Model Uncertainty Estimation 383
14.2.5 Considerations for Hybrid Model Construction 384
14.2.5.1 Hybrid Model Greyness 384
14.2.5.2 Discrepancy Hybrid Modeling 384
14.2.5.3 Machine Learning Component 385
14.2.6 Dynamic Optimization Using Hybrid Modeling 385
14.2.6.1 Optimal Feeding in Fed-Batch Bioprocesses 386
14.2.6.2 Dynamic Optimization Under Uncertainty 386
14.3 Hybrid Modeling of Microalgal Lutein Production 387
14.3.1 Introduction to Microalgal Lutein Production 387
14.3.2 Experimental Setup and Data Availability 388
14.3.3 Constructing a Hybrid Model for Microalgal Lutein Production 389
14.3.3.1 Preliminary Kinetic Modeling of Microalgal Lutein Production 389
14.3.3.2 Artificial Neural Network Implementation for Time-Varying Parameters 391
14.3.4 Dynamic Optimization of Microalgal Lutein Production 394
14.4 Summary 397
References 398
Index 407
About the Author :
Dr. Hal S. Alper is the Cockrell Family Regents Chair in Engineering #1 at The University of Texas at Austin in the McKetta Department of Chemical Engineering. His research focuses on applying and extending the approaches of metabolic engineering, synthetic biology, systems biology, and protein engineering.
