Химия 12 лекций
Towards Artificial Intelligence in Chemistry via the Rational Design of Chemical Compounds and Automated Discovery
Полынский Михаил Вячеславович
Ондар Евгения Эдуардовна
#лекции #спецкурс
Химический факультет

Курс призван отразить современную химию на языке, доступном специалистам из других областей. В этом курсе мы обсудим приложения машинного обучения к поиску новых веществ и материалов с оптимальными свойствами на основе имеющихся научных данных (рациональному дизайну), а также рассмотрим работы, в которых используются роботизированные системы для автоматического исследования новых химических реакций.

Язык проведения курса: английский

Для участия в курсе требуется знание языка на уровне (upper) intermediate или выше.

The course is designed to reflect state-of-the-art chemistry in a language accessibleto specialists from other fields. The renaissance of machine learning and the advancement ofdata science have created the possibility for rational design, the targeted design of chemicalsubstances with optimal properties using available data, without often-expensive discovery bytrial-and-error and serendipity. Intensive automation of chemical research drives us towards

the discovery of new compounds outsourced to robots (“chemputer”), as was demonstratedwith several proof-of-concept devices.

In this course, we will discuss applications of classic machine learning (ML) andartificial neural networks (ANN) in chemistry, basics of cheminformatics, chemical databases(with their applications and deficiencies). We will briefly review the latest notable research inmedicinal chemistry, catalysis, materials science, and organic synthesis to highlight theapplications of classic ML and ANN. Successful examples of automated discovery of chemicalcompounds by specialized robots will also be discussed, as well as applications of artificialintelligence in the prediction of chemical syntheses.

Jupyter Notebook, freely available Python libraries and open-source datasets will beused for interactive demonstrations. To fully understand the course, you should be able towrite simple scripts in Python using Jupyter Notebook and be familiar with NumPy and scikitlearn. Although the course is most relevant for chemists andstudents in the fields related tochemistry as materials scientists, biologists, physicists, med students, geologists, other STEMstudents can find this course interesting

МАТЕРИАЛЫ К КУРСУ: https://drive.google.com/drive...

Список всех тем лекций

Лекция 1. Introductory lecture.
The role of chemistry in the development of mankind Intuition and rationality as methods Artificial intelligence and machine learning Using Machine Learning in Chemistry Useful magazines Students' questions

Лекция 2. Python 3 and some common libraries.
Introduction Python characteristics On Anaconda and Jupiter COSMO model, Pandas and Plotly

Лекция 3. Rational design tools: Specialized libraries, Cheminformatics and Quantum Chemistry Software.
PyTorch library Avogadro Molecular Editor Jmol: Open-Source Molecular Visualizer Open Babel RDKit & Mordred Quantum-chemical methods Rendering and Compressing Molecular Movies

Лекция 4. Rational Disign Tools: Table Chemical Data, Molecular File Formats and Chemical Databases.
Data in Chemistry and Electronic Tables Xlsx and XML files JSON and SMILES InChl and .xyz

Лекция 5. Illustrated Discussion of Some Machine Learning and Data Analysis Methods.
Machine Learning Tasks and Categories Suggested Books and Courses ML Project Pipeline Key Concepts Bias-Variance Trade-off Training Artificial Networks and Deep Learning Data Analysis Practice

Лекция 6. Free Energy Relationships and Linear Regresion.
Introduction Thermochemical Equations and Thermodynamics Linear Free Energy Relationships Catalist and Catalysis Curtin-Hummet Principle and Steric Ligand Effect Model Development Workflow

Лекция 7. Chemical Structure and Molecular Descriptors.
Introduction Features and Types of Descriptors QTAIM and Electronic Effect STERIMOL Descriptors Tolman Cone Angle ECFPs Molecular Van der Waals Surface McGovan Molecular Volume Additivity of Atomic and Functional Group Properties 3D PSA and TopoPSA logP and Molecular Orbitals Molecular Dipole and Polar Environment Multipole Expansion and Molecular Graph Good Descriptor

Лекция 8. Rational Design in Medicinal Chemistry.
Reasons and Approaches Quantitative Measurement in vitro vs in vivo Lock-and-key model, Binding and ADMET Solubility and Druglikeness 2D and 3D QSAR Virtual Screening and QSAR Taste, Membrane Composition and Electronic Tongue Suggested Community

Лекция 9. Rational Design of Materials and Recognized Issues of Machine Learning in Chemistry.
Rational Design of Materials and its Descriptors Electronic Band Theory and Feature of Solids Rational Design of Battareis and Properties Prediction

Лекция 10. Rational Design of Catalysts.

Лекция 11. Predicting Organic Reactions with Machine Learning.

Лекция 12. Machine Learning in Organic Synthesis. Robots in Chemistry. Deep Learning in Drug Design.