Artificial Intelligence in Biomaterials Design and Development

1: Introduction to artificial intelligence, machine learning, and deep learning

• AI, ML, and DL: clarifying the concepts, definitions, and applicability
• Machine learning: basics
• Classification: supervised and unsupervised VS. Discrete and continuous
• Active and passive learning
• Discriminative VS. Generative models
• Classification, regression, clustering
• A brief introduction to SVM, k-mean clustering, regression and logistic regression, random forest, etc.
• Optimization (backpropagation): Gradient descent and Newton methods
• Multi-objective optimization
• Multi-fidelity optimization
• Bayesian optimization
• Dynamic programming
• Markov decision process
• Dimensionality reduction
• Multi-information source optimization

2: Useful tools and datasets for materials science and engineering

• Datasets: ZINC, ChEMBL, MOSES, AFLOW, Reaxys, SciFinder, OQMD, ICSD, NOMAD, etc.
• Software packages: Anaconda, Jupyter notebook, TensorFlow, scikit learn, RDkit, OpenChem, Chematica, Pytorch, Keras, matminer (data mining), Pymatgen, biopython JARVIS, etc.
• cloud platforms/computing: FloydHub, google cloud

3: Artificial neural networks

• Human brain inspiration
• Basic concepts: composite functions, matrix operation, backpropagation
• Recurrent neural networks (RNN)
• Long short-term memory (LSTM)
• Convolutional neural networks (CNNs)
• Graph based deep learning (graph CNN and message passing neural network)
• Graph CNNs, Weisfeiler-Lehman network
• Variational Autoencoders (VAEs)
• Generative adversarial network (GAN) and its derivatives
• multiple GANs
• Reinforcement learning (RL) and deep RL, Q-learning
• Ensemble learning

4: From human genome to materials genome

• The analogy between human and materials genome
• Chemical space, drug-like small molecules
• combinatorial chemistry
• Materials genome initiative (MGI), MI2I, NOMAD
• Genetic algorithm and evolutionary approaches for materials discovery/optimization
• Materials genotype and phenotype, mutation (inspiration from nature)
• Molecular representation: SMILE, InChI, graphs, 3D
• Simplex representation of molecular structure (SiRMS)
• Latent space and continuous representation of molecules (molecular mapping)
• Polymer genome

5: Biomaterials properties-prediction based on discriminative models

• Materials properties datasets: experimentation VS. computational quantum mechanical modelling (e.g., DFT)
• Materials descriptors and fingerprints
• Properties prediction
• Quantitative structure-activity or structure-property relationship (QSAR or QSPR)
• Training on small and big datasets
• Multi-objective learning

6: de novo materials design based on generative models

• De novo design and AI imagination
• VAE in new materials discovery
• GAN and derivatives in new materials discovery and chemical entities
• Multiple-GAN and RL in materials discovery
• Molecular structures: proposing, scoring, and optimization
• Synthesizability of molecular structures from available building blocks

7: AI-assisted synthesis planning and optimization of biomaterials

• computer-assisted synthetic planning (CASP) and computer-aided retrosynthesis
• Deep Reinforcement Learning
• Monte Carlo tree search (MCTS)
• symbolic artificial intelligence (Symbolic AI)
• Reaction condition recommendation (solvent, temperature, …)
• Prediction of reaction mechanism and reaction products
• Multiple target optimization
• Reaction products prediction (template, template-free approaches)
• chemical reactivity prediction
• Synthetic feasibility estimation

8: AI-assisted characterization of biomaterials

• In situ characterization
• AI in spectroscopic analysis (NMR, IR, XRD, …)
• AI in thermal analysis (TGA, DSC)
• AI in electrochemical analysis

9: AI-assisted evaluation of biomaterials

• Cytotoxicity and biocompatibility assessment
• Biodegradation evaluations
• Bioactivity and bioavailability
• Cell-biomaterials interactions
• Water absorption and retention
• Mechanical properties

10: AI and biomaterials in drug and vaccine development

• Traditional drug discovery
• Molecular docking
• High throughput virtual screening
• ML algorithms for drug design/discovery
• Datasets for drug design/discovery
• ML algorithms vaccine development
• ML-assisted personalized medicine (the intersection between human genome and materials genome)
• AI –based autonomous researchers/agents

11: AI and biomaterials in protein engineering

• Conformational space of protein folding (alphafold)
• Protein structure prediction
• Protein descriptors
• ML algorithms for protein engineering
• Datasets for protein engineering

12: AI in biopolymer design/discovery/engineering

• Polymer descriptors
• ML algorithms for biopolymer engineering
• Datasets for polymer design/discovery

13: AI in designing/discovery of other biomaterials

• Introduction to medicinal chemistry
• Human genome-materials genome confluence
• Bioactivities prediction
• Antimicrobial peptides
• Anticancer biomaterials
• Biomarkers development
• Immunomodulatory biomaterials
• Cell instructive biomaterials
• Crystalline biomaterials
• Electroactive biomaterials
• Stimuli-responsive biomaterials

14: AI-assisted biomaterials structures at scales

• Molecular structures prediction using AI/machine learning
• Supramolecular structure prediction using AI/machine learning
• Macroscale structure prediction using AI/machine learning

15: AI-assisted materials/scientific discoveries: Beyond pure machine learning

• Search, reasoning, and knowledge representation
• Probabilistic reasoning
• Planning
• Active learning and autonomous experiments
• Autonomous Molecular Design
• Reaction Discovery
• Autonomous researchers and robotic chemists (Bayesian experimental autonomous researcher)
• Flow chemistry, microfluidic systems, microreactors
• Human machine collaborations in materials/scientific discoveries

16: State-of-the-art and future perspectives on ML-assisted biomaterials design/discovery