Chemistry Reference
In-Depth Information
peptides. In this study, however, they were not; the few rings present in a small molecule
are needed to derive a reasonable framework. In a continuation of this work, Bemis and
Murcko focused on the various side-chains found in drugs.
[
30
]
Additional information was
included in the side-chain description, i.e. the connection point and type of framework atom
to which the side-chain was bonded. Side-chains consisting of a single (heavy) atom other
than hydrogen, e.g. chlorine, were also considered. The set of molecules extracted from the
CMC database was slightly smaller now, 5090 molecules in size. From this set, 4689 had
side-chains. The total number of side-chains was 18 664, on average four side-chains per
scaffold. The average length of a side-chain was two atoms. Side-chains of one heavy atom
in length were found most (66%). Since oxygen atoms double-bonded to a ring system
have a profound effect on the ring's electronic properties, it may be reasonable to consider
these as part of the ring. In this case, the number of side-chains was reduced to 57%.
Lameijer
et al
. explored the possibility of gaining new insights solely from the structures
that exist in the database.
[
27
]
For this, the NCI database
[
14
]
was mined. The authors reasoned
that the substructures, and the combinations they occur in, provide insight into synthetic
feasibility and 'chemical habits'. These habits emerge from an analysis of compound types
that are made frequently or substructures that are often found together. The most frequently
occurring fragments and fragment combinations were denoted 'chemical clichés'. Graph
splitting was used to break the molecules into parts suitable for mining. For this, the method
described by Bemis and Murcko
[
28
]
was adopted, with the extension that frameworks were
further split into ring systems and linkers. Another difference was that only side-chains
connected to a ring counted as side-chains. Side-chains attached to a linker were part of
the linker. Figure 8.7 shows an example of a molecule split into molecular parts according
to Lameijer
et al
.
[
27
]
HO
O
Ring systems:
O
H
BC
H
O
BC
N
O
O
Linkers:
NH
O
BC
O
NH
O
BC
NH
2
BC
O
Side chains:
O
O
BC
OH
BC
NH
2
O
Figure 8.7
Example structure (see also Figure 8.1) split into ring systems, linkers and side-
chains according to the algorithm of Lameijer
et al
.
[27]
In contrast to Figure 8.6, side-chains in
this figure stem from rings only. Side-chains connected to a linker are considered part of the
linker. Again, boxed 'B and atom type' labels are used to indicate a connection point to a ring.
