Cheminformatics with Python

1. Introduction
Focuses on the history, objectives, and scope of cheminformatics. Also, the structure of the book will be introduced.

1.1 History

1.2 Aims and Scope

1.3 Structure of the Book

Part I: Python for Cheminformatics

2. Python Basics
Introduces the basic concepts and features of the Python language and system to readers.

2.1 Installation

2.2 Python Interpreter

2.3 Variables and Data Types

2.4 Data Structures

2.5 Control Flow

2.6 Functions

2.7 Classes

2.8 Modules

2.9 Standard Library

2.10 Packages

2.11 Environments

2.12 Input and Output

2.13 Exceptions

2.14 Testing

2.15 Debugging

2.16 Profiling

2.17 Source Control

2.18 Collaboration

3. Python Packages
Provides an overview of the packages that need to be installed to conduct cheminformatics research based on Python, including scientific computing, cheminformatics, machine learning, deep learning, scientific visualization, databases, and web services.

3.1 Scientific Computing

3.1.1 NumPy

3.1.2 SciPy

3.1.3 Pandas

3.1.4 Jupyter

3.2 Cheminformatics

3.2.1 OpenBabel

3.2.2 RDKit

3.2.3 Chemfp

3.2.4 DeepChem

3.3 Machine Learning and Deep Learning

3.3.1 Scikit-learn

3.3.2 PyTorch

3.3.3 PyTorch Geometric

3.3.4 Hugging Face Transformers

3.4 Scientific Visualization

3.4.1 Matplotlib

3.4.2 Seaborn

3.4.3 PyMol

3.5 Databases

3.5.1 SQLite

3.5.2 MySQL

3.5.3 PostgreSQL

3.5.4 Redis

3.5.5 Milvus

3.6 Web Services

3.6.1 Django

3.6.2 Flask

3.6.3 FastAPI

3.6.4 OpenAPI

Part II: Data and Databases

4. Representation of Instrumental Signals
Focuses on how to represent the signals generated by the analysis instruments as vectors, matrices, tensors, etc., which can be easily imported into Python for further processing.

4.1 Vector Representation

4.1.1 Spectra

4.1.2 Chromatograms

4.1.3 Electrochemical Signals

4.2 Matrix Representation

4.2.1 Data of Hyphenated Instruments

4.2.2 Images from Single-Channel Imaging Instruments

4.2.3 Fluorescence (Phosphorescence) Spectra

4.2.4 Correlation Spectra

4.3 Tensor Representation

4.3.1 Images from Multi-Channel Imaging Instruments

4.3.2 Data of Multidimensional Instrumentation

4.4 Other Representations

4.4.1 Sparse Data

5. Representation of Molecules
Introduces common molecular formats and chemical reaction formats. Focuses on how to represent molecular structures as molecular descriptors, molecular fingerprints, molecular encodings, molecular graphs, molecular tokens, learned representations etc., which can be easily imported into Python for further processing.

5.1 Common Formats for Molecules

5.1.1 SMILES Format

5.1.2 InChI Format

5.1.3 Chemical Markup Language

5.1.4 MDL MOL Format

5.1.5 Sybyl Mol2 Format

5.1.6 XYZ Format

5.1.7 SELFIES Format

5.1.8 Crystallographic Information File

5.1.9 Protein Data Bank Format

5.1.10 FASTA Format

5.2 Common Formats for Chemical Reactions

5.2.1 CML Reaction Format

5.2.2 MDL RXN Format

5.2.3 Reaction SMILES Format

5.3 2D Representation

5.3.1 Molecular Descriptors

5.3.2 Molecular Fingerprints

5.3.3 Molecular One-Hot Encoder

5.3.4 Molecular Graphs

5.3.5 Molecular Tokenizers

5.4 3D Representation

5.4.1 Cartesian Coordinates

5.4.2 Internal Coordinates

5.4.3 3D Molecular Descriptors

5.5 Learned Representation

5.5.1 Mol2Vec

5.5.2 SMILES-BERT

5.5.3 ChemBERTa

5.5.4 Chemformer

5.5.5 MolCLR

6. Databases in Chemistry
This chapter first introduces basic database theory, history, classification, database management systems, structured query language, object-relational mapping, and database design and implementation. Then, chemical databases are classified into literature databases, spectral databases, property databases, molecular structure databases, molecular biology databases, patent databases, and standard databases. Finally, it summarizes the common chemical databases that lay the data foundation for building machine learning and deep learning models.

6.1 Basic Database Theory

6.1.1 Brief History

6.1.2 Classification

6.1.3 Database Management System

6.1.4 Structured Query Language

6.1.5 Object-Relational Mapping

6.1.6 Database Design and Implementation

6.1.7 Handling of Ultra-Large Databases

6.2 Classification of Chemical Databases

6.2.1 Literature Databases

6.2.2 Spectral Databases

6.2.3 Property Databases

6.2.4 Molecular Structure Databases

6.2.5 Molecular Biology Databases

6.2.6 Patent Databases

6.2.7 Standard Databases

6.3 Common Databases in Chemistry

6.3.1 Literature Databases of Publishers

6.3.2 Reference and Citation Databases

6.3.3 Mass Spectral Databases

6.3.4 NMR Spectroscopy Databases

6.3.5 Molecular Spectroscopy Databases

6.3.6 Atomic Spectroscopy Databases

6.3.7 Common Databases of Molecular Properties

6.3.8 Structural Databases of Small Molecules

6.3.9 Structural Databases of Macromolecules

6.3.10 Databases of Chemical Reactions

6.3.11 Common Databases in Molecular Biology

6.3.12 Google Patents

6.3.13 Standards in International Organization for Standardization

Part III: Methods

7. Instrumental Signal Processing
Presents signal processing methods for instrumental analysis by removing noises, baselines and other factors from signals through signal smoothing, baseline correction, peak detection, peak calibration, signal correction, signal derivation, signal transformation and other methods.

7.1 Smoothing Methods

7.1.1 Window Moving Average Method

7.1.2 Window Moving Polynomial Least Squares Method

7.1.3 Penalized Least Squares Method

7.2 Baseline Correction Methods

7.2.1 Improved Modified Polynomial Method

7.2.2 Adaptive Iteratively Reweighted Penalized Least Squares Method

7.2.3 Morphological Penalized Least Squares Method

7.2.4 Locally Estimated Scatterplot Smoothing Method

7.3 Peak Detection Methods

7.3.1 Common Criteria for Peak Detection

7.3.2 Peak Detection Based on Peak Properties

7.3.3 Peak Detection Incorporating Continuous Wavelet Transform

7.3.4 Multiscale Peak Detection in Wavelet Space

7.4 Peak Alignment Methods

7.4.1 Dynamic Time Warping

7.4.2 Correlation Optimized Warping

7.4.3 Peak Alignment Using Heuristic Optimization Algorithms

7.4.4 FFT Cross-Correlation for Alignment

7.4.5 Retention Time Alignment in XCMS

7.5 Signal Correction Methods

7.5.1 Multiplicative Scatter Correction

7.5.2 Orthogonal Signal Correction

7.5.3 Optical Length Estimation and Correction

7.6 Derivative Methods

7.6.1 Simple Discrete Difference

7.6.2 Window Moving Polynomial Least Squares Method

7.6.3 Fractional Order Derivation

7.7 Transformation Methods

7.7.1 Convolution and Cross-Correlation

7.7.2 Hadamard Transformation

7.7.3 Fast Fourier Transform

7.7.4 Wavelet Transform

8. Multivariate Calibration and Resolution
Multivariate calibration and resolution According to the component complexity of the systems, the multi-component systems are divided into white, grey and black systems, and the quantitative analysis methods applicable to these systems are introduced, respectively. For the white system, we introduce direct calibration and indirect calibration methods. For the grey system, the vector and matrix calibration methods are introduced. For the generalized grey system, we introduce multiple linear regression, principal component regression, partial least squares regression, model maintenance, model transformation, and variable selection methods. For the black system, iterative and non-iterative multivariate resolution methods are introduced. These methods can be used to build better multivariate calibration and multivariate discrimination models to extract accurate qualitative and quantitative information from complex systems.

8.1 Multivariate Calibration for White Analytical Systems

8.1.1 Direct Calibration Methods

8.1.2 Indirect Calibration Methods

8.1.3 Generalized Standard Addition Method

8.2 Multivariate Calibration for Grey Analytical Systems

8.2.1 Vector Calibration Methods

8.2.2 Matrix Calibration Methods

8.3 Multivariate Calibration for Generalized Grey Analytical Systems

8.3.1 Multiple Linear Regression

8.3.2 Principal Component Regression

8.3.3 Partial Least Squares Regression

8.3.4 Model Maintenance and Calibration Transfer

8.3.5 Variable Selection

8.4 Multivariate Resolution for Black Analytical Systems

8.4.1 Self-Modeling Curve Resolution

8.4.2 Iterative Target Transformation Factor Analysis

8.4.3 Evolving Factor Analysis and Related Methods

8.4.4 Heuristic Evolving Latent Projections

8.4.5 Subwindow Factor Analysis

8.4.6 Multivariate Curve Resolution Alternating Least Squares

8.4.7 Resolution by Immune Algorithms

8.4.8 Multi-Way Calibration

9. Manipulation of Molecular Structures
Focuses on how to process molecular structure information, generate molecular 3D conformations, calculate molecular descriptors, calculate molecular fingerprints, calculate similarities between molecules, perform chemical structure searches, and perform transformations or chemical reactions on molecular structures.

9.1 Working with Molecular Structures

9.1.1 Reading Molecules

9.1.2 Drawing Molecules

9.1.3 Editing Molecules

9.1.4 Writing Molecules

9.2 Conformer Generation

9.2.1 Distance Geometry

9.2.2 Experimental-Torsion Basic Knowledge Distance Geometry

9.2.3 Universal Force Field

9.2.4 Merck Molecular Force Field

9.3 Molecular Descriptor Calculation

9.3.1 Constitutional Descriptors

9.3.2 Topological Descriptors

9.3.3 Connectivity Descriptors

9.3.4 Geometric Descriptors

9.3.5 E-State Descriptors

9.3.6 Charge Descriptors

9.3.7 Kappa Shape Descriptors

9.3.8 Other Descriptors

9.4 Molecular Fingerprints

9.4.1 Morgan Fingerprint

9.4.2 MACCS Keys

9.4.3 PubChem Fingerprint

9.4.4 Daylight-like Fingerprint in RDKit

9.4.5 Atom-Pair Fingerprint

9.4.6 Topological-Torsion Fingerprint

9.4.7 OpenBabel FP Series Fingerprints

9.4.8 Mol2Vec

9.5 Molecular Similarity

9.5.1 Fingerprint Generation

9.5.2 Similarity Metrics

9.5.3 Procedure of Molecular Similarity Searching

9.5.4 Similarity Scores

9.5.5 Visualization of Similarity Contributions

9.6 Chemical Structure Searching

9.6.1 SMILES Arbitrary Target Specification

9.6.2 Full Structure Search

9.6.3 Substructure Search

9.6.4 3D Structure Search

9.7 Chemical Transformations and Reactions

9.7.1 Substructure-Based Transformations

9.7.2 Murcko Decomposition

9.7.3 RECAP and BRICS

9.7.4 Chemical Reactions Based on Reaction SMILES

9.7.5 Chemical Reactions Based on MDL RXN Files

10. Classic Machine Learning Methods
This chapter will explore classic machine learning methods for statistical inference: drawing conclusions on the chemical data at hand. These machine learning methods include pattern distances, similarity metrics, normalization, feature extraction, dimensionality reduction and visualization, clustering, classification, regression, model selection and evaluation, and other related methods.

10.1 Distance and Similarity

10.1.1 Distances in Pattern Space

10.1.2 Similarities in Pattern Space

10.2 Normalization and Feature Extraction

10.2.1 Range Scaling

10.2.2 Autoscaling

10.2.3 Normalization

10.2.4 Encoding Categories

10.2.5 Missing Value Imputation

10.3 Dimensionality Reduction and Visualization

10.3.1 Principal Component Analysis

10.3.2 Independent Component Analysis

10.3.3 Non-Negative Matrix Factorization

10.3.4 Multidimensional Scaling

10.3.5 Isometric Feature Mapping

10.3.6 Locally Linear Embedding

10.3.7 t-Distributed Stochastic Neighbor Embedding

10.3.8 Uniform Manifold Approximation and Projection

10.4 Clustering

10.4.1 K-Means

10.4.2 Hierarchical Clustering

10.4.3 Density-Based Spatial Clustering of Applications with Noise

10.4.4 Hierarchical Density-Based Spatial Clustering of Applications with Noise

10.5 Classification

10.5.1 Nearest Neighbors Classification

10.5.2 Logistic Regression

10.5.3 Linear Discriminant Analysis

10.5.4 Support Vector Classification

10.5.5 Gaussian Process Classification

10.5.6 Partial Least Squares Discriminant Analysis

10.5.7 Random Forest Classification

10.5.8 Boosting Classification

10.6 Regression

10.6.1 Nearest Neighbors Regression

10.6.2 Linear Regression

10.6.3 Ridge Regression

10.6.4 Least Absolute Shrinkage and Selection Operator

10.6.5 Elastic Net

10.6.6 Support Vector Regression

10.6.7 Gaussian Process Regression

10.6.8 Partial Least Squares Regression

10.6.9 Random Forest Regression

10.6.10 Boosting Regression

10.7 Model Selection and Evaluation

10.7.1 Subsets Split

10.7.2 Cross-Validation

10.7.3 Hyperparameter Optimization

10.7.4 Evaluation Metrics

10.7.5 Visualization

11. Deep Learning Methods
This chapter introduces deep learning methods for cheminformatics, including multilayer perceptrons, convolutional neural networks, recurrent neural networks, attention mechanisms and Transformers, graph neural networks, and generative networks.

11.1 Multilayer Perceptrons

11.1.1 Perceptron

11.1.2 Hidden Layers

11.1.3 Activation Functions

11.1.4 Optimizers

11.1.5 Forward Propagation

11.1.6 Loss Functions

11.1.7 Backpropagation

11.1.8 Training

11.1.9 Overfitting and Regularization

11.2 Convolutional Neural Networks

11.2.1 Translation Invariance

11.2.2 Cross-Correlation and Convolution

11.2.3 Convolutional Layers

11.2.4 Transposed Convolutional Layers

11.2.5 Padding and Stride

11.2.6 Pooling

11.2.7 Batch Normalization

11.2.8 Common CNN Models

11.2.9 Fully Convolutional Network and U-Net

11.2.10 Object Detection and Segmentation

11.3 Recurrent Neural Networks

11.3.1 Sequence Data

11.3.2 Vocabulary and Tokenization

11.3.3 Basic Recurrent Neural Network

11.3.4 Long Short-Term Memory

11.3.5 Gated Recurrent Units

11.3.6 Bidirectional RNN

11.3.7 Seq2Seq

11.3.8 Beam Search

11.4 Attention Mechanisms and Transformers

11.4.1 Queries, Keys, and Values

11.4.2 Attention Mechanism

11.4.3 Multi-Head Attention

11.4.4 Self-Attention

11.4.5 Positional Encoding

11.4.6 Transformer

11.4.7 Vision Transformer

11.4.8 Swin Transformer

11.5 Graph Neural Networks

11.5.1 Graph Structured Data

11.5.2 Tasks of GNNs

11.5.3 Message Passing

11.5.4 Pooling

11.5.5 Normalization

11.5.6 Readout

11.5.7 Common GNN Models

11.6 Generative Networks

11.6.1 Autoregressive Network

11.6.2 Variational Autoencoder

11.6.3 Generative Adversarial Network

11.6.4 Flow Network

11.6.5 Diffusion Model

11.7 Recent Advances in Deep Learning

11.7.1 Distributed Training

11.7.2 Hyperparameter Tuning

11.7.3 Large-Scale Pretraining

11.7.4 Contrastive Learning

11.7.5 Multimodal Machine Learning

11.7.6 Foundation Models

11.7.7 Explainable Artificial Intelligence

11.7.8 Reinforcement Learning

Part IV: Applications

12. Cheminformatics in Analytical Chemistry
In this chapter, we give a brief introduction to the application of cheminformatics in Analytical Chemistry, including data preprocessing, qualitative analysis, quantitative analysis, and structure elucidation.

12.1 Introduction

12.2 Data Preprocessing

12.3 Qualitative Analysis

12.4 Quantitative Analysis

12.5 Structure Elucidation

12.6 Summary and Outlook​​​​​​​

13. Cheminformatics in Metabonomics
In this chapter, we give a brief introduction into the application of cheminformatics in Metabonomics, including raw spectral processing, statistical analysis, functional analysis, and integrative analysis.

13.1 Introduction

13.2 Raw Spectral Processing

13.3 Statistical Analysis

13.4 Functional Analysis

13.5 Integrative Analysis

13.6 Summary and Outlook

14. Cheminformatics in Drug Discovery
In this chapter, we give a brief introduction into the application of cheminformatics in drug discovery, including data sources for drugs, targets, and diseases, druglikeness, synthetic accessibility, pharmacophore, quantitative structure–activity relationship, drug–target interaction, ADMET properties, virtual screening, molecular dynamics simulation.

14.1 Introduction

14.2 Drug Discovery Process

14.3 Data Sources for Drugs, Targets, and Diseases

14.4 Druglikeness

14.5 Synthetic Accessibility

14.6 Pharmacophore

14.7 Quantitative Structure–Activity Relationship

14.8 Drug–Target Interaction

14.9 ADMET Properties

14.10 Virtual Screening (including new paradigms such as V-SYNTHES, Deep Docking)

14.11 Molecular Dynamics Simulation

14.12 Summary and Outlook​​​​​​​

15. Cheminformatics in Materials Science
In this chapter, we give a brief introduction into the application of cheminformatics in Materials Science, including representation of materials, generative models and methods for materials, prediction of materials properties, high throughput screening, inverse design of materials.

15.1 Introduction

15.2 Representation of Materials

15.3 Generative Models and Methods for Materials

15.4 Prediction of Materials Properties

15.5 High-Throughput Screening

15.6 Inverse Design of Materials

15.7 Summary and Outlook​​​​​​​

Appendices
A: Necessary Knowledge of Mathematics
This appendix provides a rapid introduction to the basic knowledge of Mathematics that you will need to follow most of the technical content in this book. They include the basic linear algebraic operations for high-dimensional data elements, calculus to determine which direction to adjust each parameter to decrease the loss function, some basic probability for reasoning under uncertainty, and basic information theory to measure and compare how much information is present in different signals.
A.1 Linear Algebra
A.1.1 Addition and Subtraction of Vectors
A.1.2 Direction and Length of Vector
A.1.3 Scalar Multiplication of Vectors
A.1.4 Inner and Outer Products Between Vectors
A.1.5 Matrix Addition and Subtraction
A.1.6 Matrix Multiplication
A.1.7 Zero Matrix and Identity Matrix
A.1.8 Transpose of a Matrix
A.1.9 Determinant of a Matrix
A.1.10 Inverse of a Matrix
A.1.11 Orthogonal Matrix
A.1.12 Trace of a Square Matrix
A.1.13 Rank of a Matrix
A.1.14 Eigenvalues and Eigenvectors of a Matrix
A.1.15 Orthogonal Similar Transformation of a Matrix
A.1.16 Singular Value Decomposition
A.1.17 Generalized Inverse
A.1.18 Derivative of a Matrix
A.1.19 Norms of Vector and Matrix
A.1.20 Concept of Tensor
A.2 Calculus
A.2.1 Differential Calculus
A.2.2 High-Dimensional Differentiation
A.2.3 Gradients and Gradient Descent
A.2.4 Multivariate Chain Rule
A.2.5 Backpropagation
A.3 Distributions
A.3.1 Bernoulli
A.3.2 Discrete Uniform
A.3.3 Continuous Uniform
A.3.4 Binomial
A.3.5 Poisson
A.3.6 Gaussian
A.3.7 Exponential Family
A.4 Information Theory
A.4.1 Entropy
A.4.2 Mutual Information
A.4.3 Kullback–Leibler Divergence
A.4.4 Cross-Entropy

B: Editors and IDEs
This appendix introduces common Python code editors and integrated development environments.
B.1 Jupyter
B.2 Spyder
B.3 VSCode
B.4 PyCharm