reading

Cover
Dedication
Foreword
List of Contributors
Chapter 1: Introduction
1. 1.1 The Rationale for the Books
2. 1.2 The Objectives of Chemoinformatics
3. 1.3 Learning in Chemoinformatics
4. 1.4 Outline of the Book
5. 1.5 The Scope of the Book
6. 1.6 Teaching Chemoinformatics
7. References
Chapter 2: Principles of Molecular Representations
1. 2.1 Introduction
2. 2.2 Chemical Nomenclature
3. 2.3 Chemical Notations
4. 2.4 Mathematical Notations
5. 2.5 Specific Types of Chemical Structures
6. 2.6 Spatial Representation of Structures
7. 2.7 Molecular Surfaces
8. Selected Reading
9. References
10. Exercises
Chapter 3: Computer Processing of Chemical Structure Information
1. 3.1 Introduction
2. 3.2 Standard File Formats for Chemical Structure Information
3. 3.3 Input and Output of Chemical Structures
4. 3.4 Processing Constitutional Information
5. 3.5 Processing 3D Structure Information
6. 3.6 Visualization of Molecular Models
7. 3.7 Calculation of Molecular Surfaces
8. 3.8 Chemoinformatic Toolkits and Workflow Environments
9. Selected Reading
10. References
11. Exercises
Chapter 4: Representation of Chemical Reactions
1. 4.1 Introduction
2. 4.2 Reaction Equation
3. 4.3 Reaction Types
4. 4.4 Reaction Center and Reaction Mechanisms
5. 4.5 Chemical Reactivity
6. 4.6 Learning from Reaction Information
7. 4.7 Building of Reaction Databases
8. 4.8 Reaction Center Perception
9. 4.9 Reaction Classification
10. 4.10 Stereochemistry of Reactions
11. 4.11 Reaction Networks
12. Selected Reading
13. References
14. Exercises
Chapter 5: The Data
1. 5.1 Introduction
2. 5.2 Data Types
3. 5.3 Storage and Manipulation of Data
4. 5.4 Conclusions
5. Selected Reading
6. References
7. Exercises
Chapter 6: Databases and Data Sources in Chemistry
1. 6.1 Introduction
2. 6.2 Chemical Literature and Databases
3. 6.3 Major Chemical Database Systems
4. 6.4 Compound Databases
5. 6.5 Databases with Properties of Compounds
6. 6.6 Reaction Databases
7. 6.7 Bibliographic and Citation Databases
8. 6.8 Full‐Text Databases
9. 6.9 Architecture of a Structure‐Searchable Database
10. Selected Reading
11. References
12. Exercises
Chapter 7: Searching Chemical Structures
1. 7.1 Introduction
2. 7.2 Full Structure Search
3. 7.3 Substructure Search
4. 7.4 Similarity Search
5. 7.5 Three‐Dimensional Structure Search Methods
6. 7.6 Sequence Searching in Protein and Nucleic Acid Databases
7. 7.7 Summary
8. Selected Reading
9. References
10. Exercise
Chapter 8: Computational Chemistry
1. 8.1 Empirical Approaches to the Calculation of Properties
2. 8.2 Molecular Mechanics
3. 8.3 Molecular Dynamics
4. 8.4 Quantum Mechanics
Chapter 9: Modeling and Prediction of Properties (QSPR/QSAR)
Chapter 10: Calculation of Structure Descriptors
1. 10.1 Introduction
2. 10.2 Structure Descriptors for Classification and Similarity Searching
3. 10.3 Structure Descriptors for Quantitative Modeling
4. 10.4 Descriptors That Are Not Calculated from the Chemical Structure
5. 10.5 Summary and Outlook
6. Selected Reading
7. References
8. Exercises
Chapter 11: Data Analysis and Data Handling (QSPR/QSAR)
1. 11.1 Methods for Multivariate Data Analysis
2. 11.2 Artificial Neural Networks (ANNs)
3. 11.3 Deep and Shallow Neural Networks
Chapter 12: QSAR/QSPR Revisited
1. 12.1 Best Practices of QSAR Modeling
2. 12.2 The Data Science of QSAR Modeling
3. Selected Reading
4. References
5. Exercises
Chapter 13: Bioinformatics
1. 13.1 Introduction
2. 13.2 Sequence Databases
3. 13.3 Searching Sequence Databases
4. 13.4 Characterization of Protein Families
5. 13.5 Homology Modeling
6. Selected Reading
7. References
8. Exercises
Chapter 14: Future Directions
1. 14.1 Access to Chemical Information
2. 14.2 Representation of Chemical Compounds
3. 14.3 Representation of Chemical Reactions
4. 14.4 Learning from Chemical Information
5. 14.5 Training in Chemoinformatics
Answers Section
1. Chapter 2
2. Chapter 3
3. Chapter 4
4. Chapter 5
5. Chapter 6
6. Chapter 7
7. Chapter 8
8. Chapter 10
9. Chapter 11
10. Chapter 12
11. Chapter 13
Index
Substance Index
End User License Agreement

List of Tables

Chapter 2
1. Table 2.1 Different types of molecular representations of phenylalanine (without stereochemistry).
Chapter 3
1. Table 3.1 Important file formats for the representation and exchange of chemical structure information and their respective possibilities for representing or coding the constitution, the configuration, that is, the stereochemistry, and the 3D structure or conformation.
2. Table 3.2 Comparison of the SMILES and the SLN syntax (see also Section 3.2.3).
3. Table 3.3 Format details of the ATOM record (cf. Figure 3.18).
4. Table 3.4 Sample STAR‐based dictionaries.
5. Table 3.5 Collection of selected molecule editors and molecule viewers (not complete).
6. Table 3.6 Freely available toolkits and workflow environments in chemo‐ and bioinformatics.
Chapter 4
1. Table 4.1 Substituent constants for various groups to be used in Eqs. 4.1 and 4.2. σ_m for substituents in meta‐position and σ_p for substituents in para‐position.
Chapter 5
1. Table 5.1 Selected artifacts in chemistry that require complex data types.
2. Table 5.2 Example of efficiency of an ab initio geometry optimization in different coordinate sets.
3. Table 5.3 Probability (P) of collision avoidance for 32‐ and 64‐bit hash codes.
Chapter 6
1. Table 6.1 CAS databases in SciFinder [1, 4].
2. Table 6.2 Searches for references to the analgesic paracetamol and to substituted naphthalenes in SciFinder and Reaxys.
3. Table 6.3 Hits for lidocaine 6.1 in SciFinder and Reaxys with different queries.
4. Table 6.4 Search results (hits) for query structures of Figure 6.3 in SciFinder and Reaxys.
5. Table 6.5 Special patent sequence databases available from STN at cost [6].
6. Table 6.6 Free Web portals for sequence and related databases.
7. Table 6.7 Searching for compounds containing the elements Li, O, P, F, and Fe.
8. Table 6.8 3D structure databases of inorganic, organic, and biomolecules.
9. Table 6.9 Selected properties of (+)‐estrone from diverse databases.
10. Table 6.10 Major databases with thermodynamic data available at cost.
11. Table 6.11 NIST databases with thermodynamic data.
12. Table 6.12 Major spectral databases and database systems available a cost.
13. Table 6.13 Databases with special spectroscopic data at no charge.
14. Table 6.14 Single‐step reactions in Reaxys and SciFinder CASREACT (May 2017).
15. Table 6.15 Selected bibliographic databases at cost [6].
16. Table 6.16 Selected patent databases at cost.
Chapter 7
1. Table 7.1 List of some popular similarity coefficients.
2. Table 7.2 Distance distributions for structures A and B (see Figure 7.16).
c08a
1. Table 8.1 Experimental mean molecular polarizabilities and values calculated by Eq. 8.2.
2. Table 8.2 Parameters for the dependence of orbital electronegativity on charge (Eq. 8.4).
Chapter 10
1. Table 10.1 Classification of molecular descriptors by data type.
2. Table 10.2 Classification of fragment descriptors and fingerprints.
3. Table 10.3 Classification of descriptors by the dimensionality of their molecular representation.
4. Table 10.4 Invariance properties of selection of molecular descriptors.
c11a
1. Table 11.1 Basic results obtained from the application of a binary classifier to a test set with n objects, n ₁ actually belonging to class 1 and n ₂ to class 2; n ₁ + n ₂ = n. n _CA is the number of objects belonging to class C (1 or 2) and being assigned to class A (1 or 2).
c11c
1. Table 11.2 Comparison of prediction errors of the kaggle test sets using DNN [3] and BNN [22].

List of Illustrations

Chapter 1
1. Table Figure 1.1 Inductive learning.
Chapter 2
1. Figure 2.1 Hierarchy levels of molecular representation providing different information contents.
2. Figure 2.2 Classification of different types of molecular representations with examples given in the bottom line of the boxes.
3. Figure 2.3 The IUPAC name of caffeine: 1,3,7‐trimethylpurine‐2,6‐dione.
4. Figure 2.4 Scientific notation for the information on chemical elements, for example, oxygen (O).
5. Figure 2.5 Identical empirical formula of two compounds: (a) phenylalanine and (b) 2‐(3‐furanyl)‐1‐pyrrolidinecarboxaldehyde and their subdivided formula representations.
6. Figure 2.6 Different graph theoretical representations (b) to (d) of the structure diagram of 5‐ethyl‐2‐methylheptane (a). In graph theory, only the connections are important, not the length of the edges or the angles between them. (Usually, only heavy atom types are represented, not hydrogen atoms.)
7. Figure 2.7 (a) The simple graph G₁(V₁, E₁) has four vertices V₁ = {v₁, v₂, v₃, v₄} and three edges E₁ = {e₁, e₂, e₃}. The latter are all incident in vertex v₂. Thus, the degree of vertex v₂ is three, that is, d(v₂) = 3; (b) The multigraph G₂ with E₁ = {e₁, …, e₅} on the right‐hand side has an additional edge between v₂ and v₃.
8. Figure 2.8 Vertex coloring (a) and edge coloring (b) of a graph.
9. Figure 2.9 The example (a) of a simple undirected graph G₁ consists of the set of vertices V₁ = {v₁, v₂, v₃, v₄} and the sets of edges E₁ = {{v₁, v₂}, {v₂, v₃}, {v₂, v₄}}. The example (b) of a simple directed graph G₃ consists of the set of vertices V₃ = {v₁, v₂, v₃, v₄} and the sets of edges E₃ = {(v₁, v₂), (v₂, v₄), (v₃, v₂)}.
10. Figure 2.10 Graph examples of a walk (a) and a path (b) starting in v₁.
11. Figure 2.11 Examples of a subgraph and a supergraph. Graph G₂ with the vertex set V₂ = {1, 2, 3, 4, 5, 6} and graph G₃ with the vertex set V₃ = {1, 2, 6, 7, 8} are subgraphs of graph G₁ with the vertex set V₂ = {1, 2, 3, 4, 5, 6, 7, 8}. Accordingly, graph G₁ is a supergraph of the graphs G₂ und G₃.
12. Figure 2.12 An adjacency matrix of ethanal (a) is simplified stepwise by setting all matrix elements of the lower‐left corner to zero, retaining the upper‐right corner of the matrix and (b), finally, omitting the matrix elements for hydrogen atoms.
13. Figure 2.13 An incidence matrix A_I of ethanal. In the non‐square matrix, the atoms are listed in columns and the bonds in rows. Only an undirected graph is presented here.
14. Figure 2.14 Distance matrices of ethanal with (a) geometric distances in angstrom and (b) topological distances. The matrix elements of (b) result from counting the number of bonds along the shortest path between the chosen atoms (Figure 2.12).
15. Figure 2.15 A bond matrix A_B of ethanol (a) and the redundant matrix (b).
16. Figure 2.16 A BE matrix of ethanal (a) and the redundant matrix (b).
17. Figure 2.17 The BE matrix of ethanal enables the determination of the number of valence electrons on each atom (the sum of each row/column).
18. Figure 2.18 The cross sum of the BE matrix of Figure 2.17 validates the octet rule.
19. Figure 2.19 The structure diagram of ethanal represented as a connection table consisting of two lists, a list of atoms and a list of bonds. The atom list contains atom indices that have been arbitrarily assigned to each atom and are used as atom references in the bond list.
20. Figure 2.20 Connection table of ethanal.
21. Figure 2.21 More compact (nonredundant) connection table of ethanal. Only non‐hydrogen atoms are considered; bonds with the lower indices are given only once (compare with Figure 2.20).
22. Figure 2.22 Workflow for the classification of isomeric structures of organic molecules.
23. Figure 2.23 Scheme of tautomerism. X, Y, and Z can be a carbon, nitrogen, phosphor, oxygen, or sulfur atom. G represents an electrofuge or nucleofuge group, in most cases a hydrogen atom.
24. Figure 2.24 Keto–enol (a), imine–enamine (b), and lactam–lactim (c) tautomerism.
25. Figure 2.25 Thioacetic acid shows tautomerism (from left to right at the top), mesomerism (from left to right at the bottom), and ionization (top to bottom on the left‐ and right‐hand sides).
26. Figure 2.26 The substituted phenyl derivative is a typical Markush structure (a). Here, a number of compounds are described in a one‐structure diagram by different types of variability: R¹ substituent variation, R² homology variation, R³ position variation, and n frequency variation. Phenylalanine (b) is one of these structures when R¹ is COOH, R² and R³ are H. Another example that is covered by the Markush notation is (c) 3,4‐dichloro‐N‐ethyl‐α‐phenyl‐benzenebutanamine.
27. Figure 2.27 Single bonds (a) are stored as σ‐systems and double bonds (b) as π‐systems.
28. Figure 2.28 Lewis representations of (a) benzene and 1,3‐butadiene (b), as examples of resonance of aromaticity and conjugated double bonds.
29. Figure 2.29 (a) Two zwitterionic resonance structures are required for a valid VB representation of the nitro group. (b) Improper representation of a nitro group with a pentavalent nitrogen atom. (c) The hybrid diagram needs no charged resonance structures. The π‐system contains four electrons on three atoms.
30. Figure 2.30 Lone pairs, radicals, and orbitals without electrons are represented by a π‐system with two, one, or zero electrons on the corresponding atom, respectively.
31. Figure 2.31 (a) Singlet carbine and (b) triplet carbene.
32. Figure 2.32 Enol ethers (a) have two different ionization potentials (IP) (b), depending on the orbitals concerned.
33. Figure 2.33 Example of an electron‐deficient bond in diborane with a B-H-B bond.
34. Figure 2.34 The multi‐haptic bonding in organometallic complexes (e.g., ferrocene) cannot be expressed adequately by a connection table.
35. Figure 2.35 From the constitution to the configuration and then to the conformation (3D structure) of a molecule with the example of phenylalanine (from left to right).
36. Figure 2.36 Workflow for the classification stereoisomerism.
37. Figure 2.37 Butenedioic acid with (a) E configuration = fumaric acid, and (b) Z configuration = maleic acid.
38. Figure 2.38 Different possibilities for displaying stereochemistry: (a) the wedge/hashed projection is the usual way for depicting stereochemistry, (b) this representations is ambiguous and should therefore not be used, and (c) the Fischer projection is particularly used for carbohydrates.
39. Figure 2.39 The four ligands attached to a chiral center are ranked according to the CIP rules (OH > COOH > CH₃ > H); then the enantiomers of 2‐hydroxy‐propionic acid (lactic acid) can be assigned with the R (right‐hand) or S (left‐hand) configuration.
40. Figure 2.40 Examples of chiral molecules with different types of stereogenic units. (a) Lactic acid, (b) tetra‐substituted adamantane, (c) binaphthyl compound, and (d) substituted paracyclophane.
41. Figure 2.41 Cartesian coordinate system and Cartesian coordinates of chloromethane.
42. Figure 2.42 Internal coordinates of 1,2‐dichloroethane: bond lengths r₁ and r₂, bond angle α, and torsion angle τ.
43. Figure 2.43 Graphical representation of the electrostatic potential on molecular surfaces of phenylalanine: (a) dots, (b) mesh or chicken wire, (c) solid, and (d) semitransparent.
Chapter 3
1. Figure 3.1 The standard InChI of phenylalanine.
2. Figure 3.2 The standard InChI and InChIKey of phenylalanine including stereochemistry.
3. Figure 3.3 XYZ file representing ethanal.
4. Figure 3.4 Z‐matrix of 1,2‐dichloroethane.
5. Figure 3.5 All MDL file formats based on the Molfile format.
6. Figure 3.6 Structure of (2R,3E)‐4‐chlorobut‐3‐en‐2‐ol.
7. Figure 3.7 Molfile representing (2R,3E)‐4‐chlorobut‐3‐en‐2‐ol shown in Figure 3.6 in the V2000 and V3000 format. The line numbers listed in the gray columns and the descriptions are not part of a Molfile. The fundamental structure of a Molfile is indicated in the central column.
8. Figure 3.8 Detailed description of the counts line of both Molfile versions (V2000 and V3000) from the molecule shown in Figure 3.6.
9. Figure 3.9 Detailed description of the atom block of both Molfile versions (V2000 and V3000) from the molecule shown in Figure 3.6.
10. Figure 3.10 Detailed description of the bond block of both Molfile versions (V2000 and V3000) from the molecule shown in Figure 3.6.
11. Figure 3.11 Structure of sulfamide (sulfuric diamide).
12. Figure 3.12 Sample SDfile for sulfamide (sulfuric diamide) including both structural information and associated data, for example, physicochemical properties.
13. Figure 3.13 Visualization of the 3D molecular structure of α‐conotoxin PNI1 polypeptide (PDB ID: 1pen) including secondary information (as cartoons) and water molecules.
14. Figure 3.14 Title section of the analyzed PDB file. (The gray rows specify the position of the 80 possible characters and the gray column indicates the line number of the file.)
15. Figure 3.15 Remark section of the analyzed PDB file.
16. Figure 3.16 Primary structure and heterogen sections of the analyzed PDB file.
17. Figure 3.17 Crystallographic and coordinate transformation sections of the analyzed PDB file.
18. Figure 3.18 Atomic coordinate data section of the analyzed PDB file.
19. Figure 3.19 Last lines of the analyzed PDB file.
20. Figure 3.20 CML file representing (2R,3E)‐4‐chlorobut‐3‐en‐2‐ol shown in Figure 3.6. The line numbers are listed in the grey column and are not part of a CML file.
21. Figure 3.21 The JSME as an example of a molecular editor.
22. Figure 3.22 The Jmol stand‐alone application as an example of a molecular viewer.
23. Figure 3.23 Different approaches to storing and retrieving tautomeric forms of molecules – here shown with acetylacetone. In model (a) only one tautomeric form represents the molecule, whereas in approach (b) the set of all tautomeric forms is used. Representation (c) indicates that only one generalized tautomeric form describes all the tautomers of (b).
24. Figure 3.24 Standard (a) and nonstandard InChIs (b) and the InChIKeys of acetylacetone to describe mobile hydrogens in tautomeric forms. A specific tautomeric structure (c) is described by the fixed hydrogen atom layer, where one hydrogen is fixed at atoms 3 and 6. (/b4‐3 describes in addition a double bond in the stereolayer.)
25. Figure 3.25 Six different possibilities for labeling the atoms in hypochlorous acid.
26. Figure 3.26 During the iteration, the EC value of each atom is calculated by summing the EC values of the directly connected non‐hydrogen atoms of the former sphere (relaxation process).
27. Figure 3.27 The EC values of the atoms of phenylalanine (simplified to nodes for all non‐hydrogen atoms, ignoring stereochemistry and bond types) are calculated by considering the EC values of the neighboring atoms. After each relaxation process (a–d), k, the number of equivalent classes (different EC values), is determined.
28. Figure 3.28 Canonicalization starts at the atom with the highest EC value (in the example of Figure 3.27 the atom with the value of 16), which obtains the number 1.
29. Figure 3.29 The molecular graph consists of three cycles (c1, c2, c3), which can be represented by the incidence vectors of their edges. The set of three cycles is dependent, since the symmetric difference (= exclusive or) of two of the incidence vectors results in the third one. For example, the symmetric difference of c2 and c3 results in c1. Each combination of two of these cycles forms a cycle basis. Only the combination of c1 and c2 forms a minimum cycle basis.
30. Figure 3.30 Two simple cases for the reduction of a graph by removing a vertex x according to Hanser's algorithm.
31. Figure 3.31 Stepwise reduction of the number of vertices according to Hanser's algorithm. The molecular graph contains 3 rings.
32. Figure 3.32 Cubane, (a) spatial and (b) planar representation, contains six rings (c) of size four. Every combination of five of these 4‐cycle forms a valid SSSR. The SMARTS pattern [R3] matches four of the eight carbon atoms.
33. Figure 3.33 Comparison of interchangeability classes (ICs), relevant cycles (RCs), and the smallest set of smallest rings (SSSR). (a) The molecular graph contains atoms A, B, and C. (b) The molecular graph consists of six rings. (c) The membership of atoms A, B, and C to an SSSR and the set of RCs and ICs are shown in table C. All of the six rings belong to the set of RCs. Each combination of two small rings and one of the large rings forms a valid SSSR. Ring 1 and ring 2 each form a separate IC. The third IC is formed by the union of the rings 3, 4, 5, and 6.
34. Figure 3.34 (a) The cyclophane‐like molecular graph contains three URFs that are identical to the ICs. (b) Cubane contains six URFs in contrast to only one IC. (c) The number of URFs and ICs are identical for atoms A and B. Atom C is a member of three URFs but only a member of one IC.
35. Figure 3.35 Stereocenters can be identified by permutation groups. Thus, the structure is separated into a skeleton and its ligands. Both are numbered independently, by the indices of the skeleton (1–4) and those of ligands (A–D) and are described in a permutation matrix.
36. Figure 3.36 The ordered list of 24 priority sequences of the ligands A–D around a tetrahedral stereocenter. According to the number of permutations, the list can be separated into two classes (odd number of permutations on the right‐hand side or even number of permutations on the left‐hand side).
37. Figure 3.37 The permutation matrices of two structures that differ through rotation by 120°. The permutation matrix of the rotated isomer can be brought into correspondence with the permutation matrix of the reference isomer by two interchanges of two ligands (transpositions).
38. Figure 3.38 Automatic 3D structure generation.
39. Figure 3.39 Classification of automatic 3D structure generators.
40. Figure 3.40 General workflow to generate a 3D model by CORINA.
41. Figure 3.41 An unsymmetrical superphane and its superstructure.
42. Figure 3.42 The principle of longest pathways for acyclic fragments and molecules.
43. Figure 3.43 Elimination of nonbonded interactions (close contacts).
44. Figure 3.44 Superimposition of a set of conformations of 2R‐benzylsuccinate with the benzene ring fixed on the right (rotatable bonds are highlighted on the left).
45. Figure 3.45 Dependence of the potential energy curve of n‐butane on the torsion angle τ between carbon atoms C2 and C3.
46. Figure 3.46 General workflow scheme as utilized by the program ROTATE for the generation of multiple conformations.
47. Figure 3.47 Derivation of the RMS_XYZ deviation of two conformations.
48. Figure 3.48 Ring templates for a saturated six‐membered ring and a six‐membered ring with one double bond as implemented in the ring conformation table of the 3D structure generator CORINA and ROTATE.
49. Figure 3.49 Derivation of the Torsion Angle Library.
50. Figure 3.50 Derivation of a symbolic potential energy function from the torsion angle distribution of a torsion fragment.
51. Figure 3.51 The first dynamic molecular display of small molecules by Levinthal was driven by the “crystal ball.”
52. Figure 3.52 The most common molecular graphic representations of phenylalanine such as (a) wire frame, (b) capped sticks, (c) ball and stick,(d) space filling, and (e) inorganic molecules (YBa₂Cu₃O_7 − x) with balls and sticks (left) and the same molecule polyhedral (right).
53. Figure 3.53 The most common molecular graphic representations of biological molecules (lysozyme): (a) balls and sticks; (b) backbone or Cα‐stick; and (c) cartoon (including the cylinder, ribbon, and tube model).
54. Figure 3.54 Different molecular surfaces of lysozyme (5LA5.pdb): (a) van der Waals Surface, (b) Connolly surface, and (c) solvent‐accessible surface.
55. Figure 3.55 Dependence of the van der Waals energy on the distance between two non‐connected atom nuclei. With decreasing atomic distance, the energy between the two atoms becomes attraction, going through a minimum at the van der Waals distance. Then, upon a further decrease in the distance, a rapid increase in repulsion energy is observed.
56. Figure 3.56 Cross section of the 3D model of formic acid (HCOOH). The van der Waals radius of each atom of the molecule is taken and by fusing the spheres the van der Waals surface is obtained.
57. Figure 3.57 The Connolly surface is determined by moving a probe sphere (usually a water molecule) over the van der Waals surface.
58. Figure 3.58 The center of the rolling probe sphere defines the solvent‐accessible surface during movement of the probe over the van der Waals surface. Thus, the molecular surface is expanded by the radius of the solvent molecule.
59. Figure 3.59 Different isovalue‐based surfaces of phenylalanine: (a) isoelectronic density, (b) molecular orbitals (HOMO–LUMO), (c) isopotential surface, and (d) isosurface of electron cryo‐microscopic volume of the ribosome of Escherichia coli.
Chapter 4
1. Figure 4.1 A typical equation of a chemical reaction.
2. Figure 4.2 Two reaction equations showing two completely different uses for the (+) symbol: (a) giving a fully balanced single reaction and (b) combining two parallel reactions into a single equation that is also not stoichiometrically balanced.
3. Figure 4.3 The scheme of a biochemical reaction indicating not only reactant (substrate) and product but also enzyme, coenzyme, and regulator, as well as showing in which species the reaction occurs.
4. Figure 4.4 Representative, simple examples of a substitution, an addition, and an elimination reaction showing the number, n, of reaction partners, and its change, Δn, during the reaction.
5. Figure 4.5 The reaction site of an elimination reaction with the bonds to be broken crossed and the bonds made in heavy lines (a) and the three mechanisms to achieve this reaction (b–d).
6. Figure 4.6 Illustrations of the charge distribution (a), the inductive effect (b), and the resonance effect (c), the polarizability effect (d), the steric effect (e), and the stereoelectronic effect (f).
7. Figure 4.7 FMO treatment of a Diels–Alder reaction: (a) reaction equation, (b) correlation diagram, and (c) orbital coefficients.
8. Figure 4.8 The dissociation of substituted benzoic acids (x = substituent) (a) and the hydrolysis of benzoic acid methyl esters (b).
9. Figure 4.9 Input of a chemical reaction by JSME also showing the atom‐to‐atom mapping.
10. Figure 4.10 Input of a metabolic reaction by the METIS editor.
11. Figure 4.11 The present pathway of reaction information from the producer to a consumer using a reaction database indicating the many steps in newly conceiving the information.
12. Figure 4.12 The pathway of reaction information by the use of an electronic laboratory notebook (ELN) into a reaction database indicating the seamless flow of information.
13. Figure 4.13 The search for oxidations of primary alcohols to carboxylic acids (a) will obtain reaction (b) as a hit, although this reaction is in reality a hydrolysis of an ester. (c) shows the correct specification of the query to obtain reactions involving the oxidation of alcohols to carboxylic acids.
14. Figure 4.14 The reaction of formaldehyde with hydrocyanic acid to give a cyanohydrine, and the matrix representation of this reaction.
15. Figure 4.15 The reaction scheme comprising the breaking and the making of two bonds and some examples of reactions following this scheme.
16. Figure 4.16 Different levels of specification for a bond participating in a reaction.
17. Figure 4.17 The reaction scheme breaking three and making three bonds and some of the reaction types that fall into this scheme.
18. Figure 4.18 A reaction scheme that changes the number of bonds at one atom and some specific examples.
19. Figure 4.19 Consecutive application of two reaction schemes to model the oxidation of thioethers to sulfoxides.
20. Figure 4.20 Classification of reactions by the CLASSIFY approach.
21. Figure 4.21 Different classification of substituents at the reaction site.
22. Figure 4.22 Reaction center of the dataset of 120 reactions (reacting bonds are indicated by dotted lines), and some reaction instances of this dataset.
23. Figure 4.23 Distribution of the dataset of 120 reactions in the Kohonen network. (a) The neurons were patterned on the basis of intellectually assigned reaction types. (b) In addition, empty neurons were patterned on the basis of their k‐nearest neighbors.
24. Figure 4.24 Some reactions that proceed under control of stereochemistry.
25. Figure 4.25 The treatment of the stereochemistry of an S_N2 reaction by permutation group theory.
26. Figure 4.26 Part of the reaction network of the pentose phosphate pathway.
Chapter 5
1. Figure 5.1 Folding of a structure key.
2. Figure 5.2 Different usage of the (+)‐symbol in reaction representation.
3. Figure 5.3 Data model for reaction representation.
4. Figure 5.4 Raman spectra of acetaminophen [40].
5. Figure 5.5 Resolved fragment of a thin‐film spectrum of liquid N‐methylpyrrole using Lorentz functions.
6. Figure 5.6 Research pathways.
7. Figure 5.7 A simple file of the “dat” format.
8. Figure 5.8 An example of the JCAMP‐DX file format.
9. Figure 5.9 A simple PMML file.
10. Figure 5.10 Overview of the wavelet transform and multi‐resolution analysis scheme.
11. Figure 5.11 An initial dataset has to be split into two or three parts.
Chapter 6
1. Figure 6.1 Lidocaine 6.1, an important local anesthetic drug and related structures (see Table 6.3).
2. Figure 6.2 Markush‐type query.
3. Figure 6.3 Keto–enol tautomerism of acetylacetone.
4. Figure 6.4 Representation of π‐complexes in chemistry (right formulas) and in databases (left formulas).
5. Figure 6.5 (+)‐Estrone.
6. Figure 6.6 Chemistry‐aware three‐tiered architecture.
Chapter 7
1. Figure 7.1 Different types of representation of the chemical graph of 4‐methylcyclohexene (note that the numbering does not conform to the IUPAC rules): (a) labeled graph (2D – arbitrary atom labeling); (b) connection table (CT); (c) linear notations; (d)topological indices; and (e) registry numbers.
2. Figure 7.2 Basic principle for efficient full structure search (unique and semi‐unique techniques).
3. Figure 7.3 Hashing function and fast address lookup.
4. Figure 7.4 Mappings between the query graph (G_Q) and the target graph (G_T). The notation (2, 1, 3, 4) means that atom 1 of the query subgraph G_Q is mapped to atom 2 from the target graph G_T, query atom 2 to target atom 1, query atom 3 to target atom 3, and query atom 4 to target atom 4.
5. Figure 7.5 Search tree of mappings obtained by applying the backtracking algorithm to the pair of structures G_Q and G_T (see the graphs in Figure 7.4). Array (M₁, M₂, M₃, M₄) denotes the mapping 1 (of G_T) → M₁ (of G_Q), 2 → M₂, 3 → M₃, 4 → M₄.
6. Figure 7.6 Backtracking approach realized as depth‐first search algorithm. Dotted arrows trace the route used for traversing all mappings in the search tree. Each node in the tree corresponds to a mapping between G_Q and G_T.
7. Figure 7.7 Screening process as part of DB substructure search: (1) calculate query keys, (2) filtrate DB by comparing the query bits with the pre‐calculated bits, and (3) use the screening result list for the final mapping.
8. Figure 7.8 Generation of hashed fingerprints.
9. Figure 7.9 Classes (groups) of topologically equivalent atoms. The bold numbers are the atom labels; the numbers in parentheses are the equivalence classes.
10. Figure 7.10 Maximum common substructure (MCS) of two chemical structures.
11. Figure 7.11 SMARTS queries examples.
12. Figure 7.12 The similarity search process.
13. Figure 7.13 The similarity concept applied in different chemical spaces.
14. Figure 7.14 Descriptor examples (0D, 1D, 2D, 3D) for the definition of similarity measures.
15. Figure 7.15 Flowchart of a pharmacophore mapping algorithm. Antihistamine pharmacophore [55] is searched against target molecule. The pharmacophore is defined by 3 features: two aromatic rings (A) and amino group (N).
16. Figure 7.16 Structures A and B with corresponding distance matrices (distances are given in nanometers).
17. Figure 7.17 Examples for DNA score matrices for the calculation of the “edit distance” similarity metric: (a) the simplest scoring matrix that corresponds to the Hamming distance and (b) the customized scoring matrix that distinguishes different types of DNA changes.
18. Figure 7.18 The stages of the DP algorithm: (a) forward filling of the scores S(i,j) and (b) backtracking to obtain the optimal alignment(s).
19. Figure 7.19 Flowchart of a heuristics‐based algorithm for fast sequence alignment.
Chapter 9
1. Figure 9.1 An indirect approach to the prediction of properties.
Chapter 10
1. Figure 10.1 Fragmentation of a molecule; the fragment code counts the number of occurrences of a fragment.
2. Figure 10.2 The derivation of fingerprints.
3. Figure 10.3 Screen of ChemoTyper: on the left is part of the Chemotype library; on the right is part of a dataset of 8193 organic molecules with the fitting chemotypes highlighted.
4. Figure 10.4 Screen of ChemoTyper: on the left are three chemotypes of the thalidomide skeleton with chemotypes differentiated by sigma charge (green) and total charge (blue); on the right‐hand side is part of a dataset that indicates hits for the two different chemotypes.
5. Figure 10.5 MNA descriptors for the atoms of phenol of levels 0, 1, and 2.
6. Figure 10.6 Self‐organizing map of bonds of a dataset of organic molecules. Black neurons (center) reactive bonds; gray neurons, non‐classified bonds; light gray, non‐reactive bonds; cross, contain both reactive and non‐reactive bonds; white, empty neurons (no bonds mapped).
7. Figure 10.7 Hydrogen‐depleted graph of 2,2‐dimethylbutane (2).
8. Figure 10.8 Calculation of the ECFP descriptors starting from an atom.
9. Figure 10.9 Scheme of an electron diffraction experiment.
10. Figure 10.10 Procedure for encoding a structure with an RDF code.
11. Figure 10.11 Radial distribution function of an organic molecule indicating how the longer C-S bond expresses itself in the peak for 1–2 and the peak for 1–3 interactions.
12. Figure 10.12 Comparison of the RDF code for aromatic compounds with different substitution patterns.
13. Figure 10.13 Comparison of the radial distribution function of the chair, boat, and twist conformations of cyclohexane (hydrogen atoms are not considered).
14. Figure 10.14 (a) Example of a chiral molecule. (b) The atoms A, B, C, and D are those directly bonded to the chiral center. The neighborhood of atom A is the set of atoms whose distance (in number of bonds) to A is less than their distance to B, C, and D. (c) Example of a combination of four atoms (i, j, k, and l), each at a different ligand (A, B, C, or D) of the chiral center.
15. Figure 10.15 Graphical representation of f_CICC(u) versus u for (+) (R) and (−) (S)‐2‐Pyrrolidinemethanol sampled at 75 evenly distributed points between −0.03 e² Å⁻¹ and +0.03 e² Å⁻¹. Hydrogen atoms not bonded to chiral carbon atoms were not considered.
16. Figure 10.16 Mapping the coordinates of points on a molecular surface into a self‐organizing neural (Kohonen) network.
17. Figure 10.17 The surface of a torus, a plane without beginning and end, is stretched out into two dimensions by making two perpendicular cuts.
18. Figure 10.18 Self‐organizing map of 2‐chloro‐4‐hydroxy‐2‐methylbutane colored by the molecular electrostatic potential.
19. Figure 10.19 A self‐organizing map (SOM) is trained with points on the molecular surface of n‐butane (for visualization the surface of the four carbon atoms and their hydrogen atoms are colored differently). This SOM is now taken as a template, and points from the surface of 1‐propanol are sent through this SOM, indicating the two hydrogen atoms missing in 1‐propanol against n‐butane as empty (white) neurons.
20. Figure 10.20 An SOM trained with n‐butane is taken as a template for various 1‐substituted propane derivatives indicating the differences in the molecular surface by empty (white) neurons.
21. Figure 10.21 The box containing a steroid indicating the points for calculating the steric or electrostatic field in the CoMFA approach.
22. Figure 10.22 Extension of the QSAR method by descriptors not based on structure.
23. Figure 10.23 Structure descriptors contain information from physics (geometric resolution), chemistry (properties of atoms), and mathematics (mathematical transformation).
24. Figure 10.24 The comparison of a molecule with a human.
Chapter 12
1. Figure 12.1 QSAR modeling workflow. This workflow is implemented within the Chembench portal (https://chembench.mml.unc.edu/; see also Ref. [18]).
2. Figure 12.2 Data cycle in cheminformatics.
3. Figure 12.3 General workflow for comprehensive curation of chemogenomics datasets.
4. Figure 12.4 Example of duplicate retrieval using PubChem ID, SMILES, chemical names, InChI keys (InChI strings are not shown but are also different), and 2D similarity. Note that the computing 2D similarity as Tanimoto coefficient using CDK descriptors yields T_c = 1 (implicating structural duplicates) for the two curated compounds (no salts, standardized functional groups and aromatization), whereas all other representations fail to suggest that the two compounds are identical.
Chapter 13
1. Figure 13.1 Growth of GenBank from 1982 to 2016.
2. Figure 13.2 Sequencing cost per genome (3 billion base pairs) from 2001 to 2015. Please note the logarithmic scaling of the y‐axis.
3. Figure 13.3 Excerpt from GenBank entry AF123456 of a messenger RNA coding for a transcription factor in chicken (Gallus gallus). Dots […] denote deleted lines.
4. Figure 13.4 Excerpt from a pairwise sequence alignment of Abl (ABL1_HUMAN) and Src (SRC_HUMAN) tyrosine kinases illustrating the meaning of identity, similarity, and gaps.
5. Figure 13.5 Degree of sequence identity as a function of the point accepted mutations (PAM) during protein evolution.
6. Figure 13.6 BLOSUM62 scoring matrix. Amino acid exchanges that result in positive scores are highlighted in gray. The left column gives the three‐letter amino acid code in addition to the one‐letter code used for the labeling of the scoring matrix.
7. Figure 13.7 Example of a hit obtained from an NCBI BLAST sequence search against the UniProtKB/Swiss‐Prot database using the sequence of Src (SRC_HUMAN) as a query.
8. Figure 13.8 Steps of the progressive multiple sequence alignment.
9. Figure 13.9 (a) Section from a multiple sequence alignment of eight Src‐family tyrosine kinases (Src, Lck, Hck, Lyn, BLK, Fyn, Yes, Fgr) and the Abl1‐kinase. The sequence positions refer to SRC_HUMAN. A “*” marks a strictly conserved sequence position; “:” and “.” denote decreasing degrees of sequence similarity. The three strictly conserved residues, G277, G279, and K298, which are part of the ATP‐binding site, are marked in bold/italic. (b) Location of residues G277, G279, and K298 in the three‐dimensional structure of Src kinase (PDB entry: 4MXO [25]). The conserved residues are shown as a space‐filled presentation, and an inhibitor is shown as a sticks model. Figure prepared with the UCSF Chimera package [26].
10. Figure 13.10 Excerpt from the PROSITE entry PS00107 describing the protein kinase ATP‐binding region signature.
11. Figure 13.11 Overlay of an Src kinase homology model (black) with the Src kinase crystal structure (PDB: 4MXO [25]). The model was generated based on the crystal structure of Abl kinase (PDB: 4WA9 [38]) sharing 47% sequence identity to Src for the modeled region. Figure prepared with the UCSF Chimera package [26].

Chemoinformatics

Basic Concepts and Methods

Copyright

Dedication