Chemistry Reference
In-Depth Information
of the 2D structure of a molecule. It is the result of fragmentation of the molecule according
to some 'breaking rules'. This chapter focuses on the
computational fragment
. We review
fragment discovery and evaluation in the context of large molecular databases as described
in the recent literature. Definitions, use and applications of fragments are addressed in
addition to fragmentation methods. Fragmentation of 3D molecular structures will not be
discussed.
[
4
]
In Section 8.2, we discuss the ways in which fragments can be derived. In the
Section 8.3, a few examples of what can be learned from such fragmentation methods are
presented together with their applications.
8.2 Fragmentation Methods
What is considered a fragment depends on the definition. A 'ring' could be a fragment or
a particular chain of carbon atoms could be a fragment. The definition follows from the
breaking rules that are used. To find structural patterns in a database, molecules should be
broken into manageable parts that are readily analyzed. Graph theory is extensively used to
this end (see Section 8.2.1). There are two approaches to molecule fragmentation. The first
approach is to find all possible fragments that form some part of the molecular structure;
the second is to dissect the molecule into fragments according to predefined (breaking)
rules. The first approach allows a complete analysis of the fragments that exist in the set.
However, the number of substructures for a single structure may then become very large,
even for amoderately sizedmolecule. Several methods allow consideration of all (potential)
fragments for analysis without generation of the full substructure set. The substructure
approach will be the subject of the subsections
Frequent subgraph mining
and
Common
substructures
in Section 8.2.2. The second fragmentation approach generally has a lower
yield of fragments per molecule. Fragments result from 'breaking' the molecular structure
into nonoverlapping, predefined parts. Thus, 'ring structures'may be defined in addition to
functional groups. Fragmentation into molecular building blocks according to predefined
rules follows in the subsections
Molecular building blocks
and
Virtual retro-synthesis
in
Section 8.2.3.
8.2.1 Graph Representation
Graph theory plays an important role in fragmentation. The 2D structure of a molecule and
its fragments are often represented as graphs.
[
5
]
Agraph is amathematical object that consists
of a set of vertices or nodes and a set of edges that connect these nodes. The molecular
structure conveniently translates into a graph, where vertices represent the atoms and edges
represent the bonds.
[
5
]
This abstraction enables the use of generic methods that are under
study in graph theory, such as the discovery of rings (cycles).
To illustrate the representation of molecules as graph, let us consider the sample struc-
ture in Figure 8.1 (taken from the PubChem compound database,
[
1
]
accession number
CID9959891). Figure 8.2 shows the graph representation of the molecule in Figure 8.1
Hydrogen atoms, even when connected to heteroatoms, are omitted. Note that with stand-
ard graphs, representation of the molecule is limited to reproducing the connection pattern
(connectivity) between the atoms. Any other information such as atom type or bond order
is disregarded.
