The document discusses cheminformatics, emphasizing its role in chemical information management and retrieval for drug discovery and optimization. It details techniques such as structure searching, molecular similarity assessments, and representations of chemical structures including SMILES and InChI. Additionally, it addresses various databases used in cheminformatics, methods for comparing molecular structures, and the importance of molecular similarity in predicting biological activity.

1. CHEMINFORMATICS
Subhasis Banerjee (Asst. Professor)
(Gupta College of Technological Sciences, West Bengal.

2. Contents
•An overview on CHEMINFORMATICS
•Searching Chemical Structures
•Structure-spectra correlation

3. Chemoinformatics encompasses the design, creation, organisation, management,
retrieval, analysis, dissemination, visualization and use of chemical information
The mixing of information resources to transform data into information and
information into knowledge, for the intended purpose of making decisions faster in
the arena of drug lead identification and optimization

4. Concepts and techniques:
Structure search (Various Databases)
Molecular similarity
Substructure search
Quantitative-structure activity relationships (QSAR)
3D structure generation from a 2D structure
Conformer generation
Algorithms

5.  How can a molecular structure be stored on a
computer?
 Common names: aspirin
 IUPAC name: 2-acetoxybenzoic acid
 Formula: C9H8O4
 As an image (PNG, GIF, etc.)
 CAS number: 50-78-2
 File format: ChemDraw file, MOL file, etc.
 SMILES string: O=C(Oc1ccccc1C(=O)O)C
 Binary Fingerprint: 10000100000001100000100100000001

6.  Simplified Molecular Input Line Entry System
 Weininger, J Chem Inf Comput Sci, 1988, 28, 31
 Examples:
 CC represents CH3CH3 (ethane)
 CC(=O)O represents CH3COOH (acetic acid)
 Basic guidelines:
 Hydrogens are implicit
 Parentheses indicate branches
 Each atom is connected to the preceding atom to its left
(excluding branches in-between)
 Single bonds are implicit, = for double, # for triple
 What is C(C)(C)(C)C?

7.  International Chemical Identifier
 Line notation developed by NIST(National Institute of
Standards and Technology) and IUPAC
 Goal: An index for uniquely identifying a molecule
 Aspirin: InChI=1/C9H8O4/c1-6(10)13-8-5-3-2-4-7(8)9(11)12/h2-
5H,1H3,(H,11,12)/f/h11H
 Features
 Derived from the structure (unlike CAS number)
 One-to-one relationship between InChI and structure
 Notes
 Not human readable or writeable
InChI

8. 8
 World Drug Index (WDI)
 Available Chemical Directory (ACD)
 National Cancer Institute (NCI)
 MayBridge (3D formatted database, by the company)
 Beilstein (the largest database for organic chemistry since 1771, now
maintained by Elsevier)
 ZINC (UCSF) (3D formatted database, by UCSF, since 2005)
 World of Molecular Bioactivity (WOMBAT)
 Toxic Substance Control Act (TSCA)
 Medchem
 Zinc
 CSD
 PDB

9. 9

10. 10

11. 11

12. 12

13. 13

14. 14

15. 15

16. 16

17. 17

18.  X-ray crystal structures of more than 250,000
compounds (organic and organometallic)
 Established in 1965
 Textual queries
 Structural queries
 Specific 3D constraints (conformation or distance
variables)

19.  More than 25,000 X-ray and NMR structures of protein and
protein-ligand complexes
 Some nucleic acid and carbohydrate structures
 Established in 1971 at Brookhaven National Laboratory; now run
by a consortium
 Retrieval by textual queries or in some interfaces by amino acid
sequences

21.  Full structure search; query is complete molecule
 Substructure search; query is a pattern of atoms and bonds
 Superstructure search; query is a complete molecule

22.  Many databases contain millions of structures, so search speed is
important
 Simplest approaches uses canonical representation for query and
database structures (e.g. canonical SMILES)
 could sort database SMILES into alphanumerical order
 search sorted list for match with query
 “Hash table” lookup can improve search speed
 calculate hash-code (“idiot number” in predefined range) from SMILES for
each database structure
 this is address (disk file or memory) at which full representation is stored
 only SMILES which have same hash code need to be compared

23.  New compounds need to be added regularly
 used to be done by chemical information specialists
 now frequently done directly by bench chemists
 Registration system must
 check consistency of input data
 e.g. compare molecular formula with structure
 check that compound is really new
 different ways of handling tautomers, salts, stereoisomers etc.
 assign registry number
 add supplementary data (melting point etc.)
 make data immediately available for search

24. In graph theory terms, when two full structures match, their graphs
are said to be isomorphic
 each node N1 in G1 must be mapped to a node N2 in G2
 neighbours of N must map to neighbours of N

25.  so far we’ve considered matching one query substructure
against one database full structure
 each structure from the database needs to be compared against
the query in turn
 many will fail because they don’t contain the query
substructure
 “screening” allows many of these to be eliminated before
we get to this stage

26.  The fragments present in a structure can be represented as a
sequence of 0s and 1s
00010100010101000101010011110100
 0 means fragment is not present in structure
 1 means fragment is present in structure (perhaps multiple
times)
 Each 0 or 1 can be represented as a single bit in the computer
(a “bitstring”)
 For chemical structures often called structure “fingerprints”

27.  Build a fingerprint for the query substructure
 Only those database structures that contain all the fragments in the
query can possibly match the query
Query: 00000100010101000001010011010100
DB struct 1: 00010100010101000101010011110100 MATCH
DB struct 2: 00000000100101001001000011100000 NO MATCH
 Comparing fingerprint bitstrings is very fast (logical & operation)
 Only those structures that pass the screening stage need to be
considered as candidates for atom-by-atom isomorphism search

28.  Ideally we want to eliminate as many structures as
possible at the screening stage
 99% screenout or more would be good
 Fingerprint construction can help in this
 frequency distributions of fragments in a large
database are very “skewed”
 A few fragments occur in almost all compounds
will therefore give little or no screenout
 Many fragments occur in very few compounds
need very long fingerprint (lots of fragments) to ensure that we will
have some in the query

29.  Do the time-consuming work in advance
 Full structure search provides an example of this
 Canonicalisation is a slow process (NP-complete)
 but it can be done in a pre-processing of the file, independently
of the query
 then store the canonical representations
 can do rapid matches against a canonicalised query structure
 this is faster than using a graph isomorphism algorithm on non-
canonical representations

30.  Similar principles are used in some substructure search systems
 A tree structure is built, classifying all the atoms found in all the
structures in the database
 first level based on atom type
 second level based on number of connections
 third level based on type of first neighbour
 fourth level based on type of second neighbour etc.
 lower levels based on classifications applied to neighbours (relaxation)
 bottom of tree lists structures that contain this class of atom

31. C — C — B rC — C — F
C — C C — O
C — CC — B r
C — — C — |
— C —
|
—C—
|
C
C — C — F
|
C
C — C — F
|
F
C e n t r a l
a t o m t y p e
N u m b e r o f
C o n n e c t i o n s
F i r s t
N e i g h b o u r
S e c o n d
N e i g h b o u r
T h i r d
N e i g h b o u r

32.  Search can be done by tracing tree, looking for atom classes found
in query
 combine lists of structures found at the bottom
 A backtracking atom-by-atom search may be needed to check hits
found
 Best-known example is Beilstein’s Crossfire
 Main problem is updating the trees when new structures are added
to the database

33.  Queries for substructure search systems may be more complicated
than simple subgraphs
 Different systems provide different capabilities
 variable atom and bond types
 specification of allowed substitution
V a r ia b le
a t o m t y p e
C (s0)
C (s1)
[F,Cl,Br,I]
s in g le o r
d o u b le b o n d
n o f u r t h e r
s u b s t it u e n t s
o n e o r m o r e
s u b s t it u e n t s

34.  Daylight uses an extension of SMILES to describe
structure queries (SMARTS)
 can attach various properties to each atom
 [CX3] carbon with 3 connections
 [Nr5] nitrogen in a ring of size 5
 properties can be combined with logical operators
 ! (NOT)
 & (AND – high precedence)
 , (OR)
 ; (AND – low precedence)

35.  complex patterns can be specified this way:
 [F,Cl, Br, I] any of the halogen atoms
 [!C;!R0] heteroatom in a ring
 $(smarts_string) can also be used as an
atom property
 this is called recursive SMARTS
 e.g. $(NC=*)
 nitrogen single-bond carbon double-bond any-atom
 (i.e. an amide)

36.  recursive SMARTS can be used to describe very
complex patterns
 e.g. primary or secondary amine, but not amide
[ N & X 3 ; H 2 , H 1 ; ! $ ( N C = * ) ]
n i t r o g e n
3 c o n n s
n i t r o g e n
c o n n e c t e d t o
c a r b o n w i t h d o u b l e
b o n d t o a n y a t o m
A N D A N D O R
2 a t t a c h e d
h y d r o g e n s
1 a t t a c h e d
h y d r o g e n
A N D N O T

37.  Searching Markush structures in patents
 nature and origin of Markush structures
 fragment codes
 topological systems (MARPAT, Markush DARC)
 Reaction searching
 atom-atom mapping
 Maximal Common Substructure search
 what is the largest substructure common to two molecules?

38.  Data mining for conformational properties and intermolecular
interactions (CSD & PDB)
 Data mining for information about intermolecular interactions
(CSD & PDB)
 Further understanding of the nature of protein structure and its
relationship to amino acid sequence (PDB)
 Homology modeling (comparative modeling) (PDB)

39. O
O N
a
b
c
a = 8.62 0.58 Angstroms
b = 7.08 0.56 Angstroms
c = 3.35 0.65 Angstroms+-
+-
+-
O
O
O
O
OO
N
O
O
O
N
N
N O
O
O
O
O
O
N
N
N
N
S
O
O
O O P O
O
O P O
O P
O
O
O
O
N
N
N
N
N
O
O
O O
O
N
N
N
O
N
O
O
O

40.  Instead of searching for all molecules containing a given
substructure, we search for molecules “similar” to a
given target molecule
 Similar property principle:
“Structurally similar molecules are expected to exhibit
similar properties or biological activities”
Mark Johnson and Gerry Maggiora (Eds.) Concepts and Applications of
Molecular Similarity. Wiley, New York, 1990

41. “Similarity is in the eye of the beholder”
Similarity can be measured in many different ways
 equivalence classes
 can say that two molecules are similar, or that they are different
 numerical measures
 can say that two molecules have a similarity of, e.g. 0.85
 similarity coefficients usually have values between
0.0 (totally different) and 1.0 (identical)
 distance measures
 “opposite” of similarity (0.0 = identical; may have no maximum, or
may be normalised to fix maximum limit)

42.  Two different molecules with the same graph
 if node and edge labels are ignored
N
OH
O
N
CH3
CH3

43.  Normally calculate some numerical measure of similarity
between molecules
 Query structure is a “target” molecule
 Database structures can be ranked in decreasing order of
similarity to target
 find all molecules with > threshold similarity to target
 find N most similar molecules to target
 No particular substructure is required in the retrieved
molecules
 but they will have structural features in common with target

44.  Similarity measures are most commonly calculated from
structure fingerprints
 count the bits that are “on” in both molecules
 count the bits that are “on” in each molecule separately
struct A: 00010100010101000101010011110100 13 bits on (A)
struct B: 00000000100101001001000011100000 8 bits on (B)
AAND B: 00000000000101000001000011100000 6 bits on (C)
 similarity coefficient can
be calculated from A, B and C
A
B
C

45.  similarity = C
A + B – C
= 6 / (13 + 8 – 6) = 0.4
 The number of bits set in both molecules divided by the
number of bits set in either molecule
 The Tanimoto coefficient is the most commonly used
similarity coefficient in chemical informatics
 also called the Jaccard coefficient
A
B
C

46.  similarity = 2C
A + B
= 12 / (13 + 8) = 0.57
 Does not give the same values as the Tanimoto
coefficient, but will rank molecules in the same order of
similarity to a target
 i.e. “monotonic” with the Tanimoto coefficient
 Also called the Czekanowski or Sørenson coefficient
A
B
C

47.  similarity = C
√ (A × B)
= 6 / √(13 × 8) = 0.588
 Not monotonic with the Tanimoto and Dice coefficients,
but highly correlated with them
 also called the Ochiai coefficient
A
B
C

48.  The three coefficients discussed so far all ignore bits that
are off in both molecules
 is common absence of features evidence of similarity between
them?
 are a camel, a horse and a nematode similar because they all
lack wings?
 are a bat, a heron and a dragonfly similar because they all have
wings?

49.  Are these molecules similar because they all lack
heteroatoms?
CH
CH
C
H
CH
CH
C
H
CH2
CH2
CH2
CH2
CH2
CH2
CH2
CH2
CH2
CH2
CH2

50.  A similarity coefficient that takes into account the
fingerprint bits that are off in both molecules (D)
 similarity = C + D
N
= (6 + 17) / 32 = 0.719
 N is length of fingerprint
 N = A + B – C + D
A B
CD

51.  This provides a means to substructure similarity
search
 also possible with maximal common subgraphs
 A and B could be number of atoms in each molecule,
and C could be number of atoms in their maximal
common substructure
 fingerprint-based similarity is generally faster than
identifying MCS
 but common features (fragments) will be smaller

52.  Many different fragment types have been used for
generating fingerprints for use in similarity searching
 atom sequence
 linear path of atoms and bonds through molecule
 may generate only paths of certain lengths
 augmented atom
 atom and its immediate neighbours

53.  Ring composition
 atom/bond sequence around a ring
 question of which rings to choose
 Ring fusion patterns
 sequence of ring connectivities around a ring
 for each atom specify number of ring bonds it has
 Atom pairs
 pair of atoms in same molecule, with number of bonds in shortest path
between them
 additional differentiation between atom types
 number of attached hydrogens / pi-bonds
 Topological torsions
 connected sequence of 4 atom types
 atom types as described for atom pairs

54.  Sometimes specific fragments (with detailed description
of atom and bond types) are too specific to be of much
use in fingerprints
 very low frequency ⇒ very sparse fingerprints
 Atom and bond types can be generalised
 any ring bond
 any halogen (F, Cl, Br, I)
 any chalcogen (O, S, Se, Te)
 This gives fragments with higher frequency

55.  fragments can be used to describe the 3D
structure of a molecule too
 usually involve interatomic distances and/or bond
angles
 because distance values are continuous variables, they
are “binned”
 each bin represents a range of distances
 e.g. distance of 3.000 – 3.999 Å
 each bin corresponds to a fingerprint bit

56.  a popular 3D descriptor is the 3-point pharmacophore
 molecule is analysed to identify “pharmacophoric points”
 points in molecule likely to be involved in binding to a
receptor site
 positive charges
 negative charges
 hydrogen-bond donors (e.g. –OH, – NH2)
 hydrogen-bond acceptors (e.g. =O)
 aromatic groups
 hydrophobic groups
 pharmacophoric points do not necessarily coincide with the
positions of individual atoms

57.  each fragment consists of 3 pharmacophoric points
 the distances between
each pair of these points
are binned and used to
set fingerprint bits
 4-point pharmacophore fragments are also used
 Different people have used slightly different definitions
of pharmacophoric points
H B D
A r y
H B A2 . 8 Å
3 . 6 Å
1 . 2 Å

58.  how do we decide which descriptors and similarity or
distance measures are “best”?
 go back to “similar property” principle:
“structurally similar molecules are expected to exhibit similar
properties or biological activities”
 we can do some experiments using
 various different sets of descriptors
 dataset of compounds with known biological activity or
measured physico-chemical property value

59.  Problem may be that 3D descriptors we have (3-point
pharmacophores with binned distances) are not good
enough
 there may be “spurious accuracy” in the detailed distances
involved
 conformational flexibility may be causing problems (as one
distance gets larger, another gets smaller)
 molecule may change conformation during binding
 some improved success has been found by identifying
“projection points” for hydrogen bond donors/acceptors (i.e.
where they’re pointing to, not where they are)

60.  2D descriptors provide “bounds” on possible 3D
conformations
 “2½D” descriptors (including some stereochemical
information) may be useful
 “Superiority” of 2D descriptors in some studies may be
artifact of datasets used
 datasets may have large numbers of close analogues
 these will have high 2D similarity, as well as correlated activity

61. Structure Spectra Correlation

62. Spectroscopy:
It all have to tell about the interaction of electromagnetic radiation with matter
How it HELPS?
With the sophisticated machineries, viz., UV, IR, NMR, MASS and many more,
it unveil the real chemical feature of a compound of any values

64. Thank you