As far as biological activity is concerned, chemical space is as empty as the celestial
variety. However, the chemical space around privileged structures is unusually rich in
biological activity. Minor changes to privileged structure molecular scaffolds allow
unrelated receptors to be modulated. Small-molecule privileged structures, such as the
benzodiazepine scaffold, have been known and exploited for some time. However, the
pharmaceutical industry is currently facing substantial increases in development costs
and decreases in productivity. These indicate that a different drug design and
development approach or at least a different toolkit is required. The mechanism by
which small molecule and protein privileged structures acquire their remarkable facility
was investigated in this work, through research at the intersection of medicinal
chemistry, chemoinformatics, combinatorial mathematics and high performance
Chemoinformatics and Computational Biology have aided our understanding of the
chemistry that underpins and creates biology. Biochemistry, as a by-product of
evolution, reuses components and is inherently modular. Nature uses and reuses
privileged structure scaffolds to create ligands that can modulate the activity of different
The diverse receptors upon which privileged structure-containing ligands can act may
share a common physicochemistry. Mimicking the common physicochemistry of
protein binding sites would therefore be a productive approach for the development of
drugs. The common features of diverse ligand binding sites for all proteins in the
Protein Data Bank (PDB) were analysed to identify and describe these privileged
We were able to identify over eight hundred different common protein surface features
in proteins' ligand binding sites. The features were biologically relevant (we were for
example able de novo to identify the binding motif for the benzamidine protease
inhibitor) and features were found with both diverse and conserved sequences, folds,
functions and ligands. A lesser explored area of privileged structures are protein-privileged structures, for
example those comprising the toxic components of the venoms of cone snails.
Although protein-protein interactions are generally considered undruggable, nature has
demonstrated that disulphide bond-rich peptides can mimic protein-protein interfaces.
Disulphide bond placements and arrangements prescribe these molecules’ shapes,
making them good molecular scaffolds as their sequences can be substantially modified
without altering their structure. To facilitate the development of peptide
proteomimetics, the arrangement and diversity of disulphide bonding patterns were
investigated and compared. Analysis of this class of molecular scaffolds will allow
drugs to be developed to modulate protein-protein interactions – a biological target that
historically was too large and too complicated to develop drugs for. This work analysed
the properties of these disulphide bond-rich privileged molecular scaffolds, and
developed tools to facilitate their use in developing drugs.
The analyses of disulphide bond-rich protein scaffolds identified two distinct
populations of proteins. Partitioning resolved three unexpected results found in
unpartitioned analyses. The two populations had different headers, folds, bond
arrangements, chain lengths, and different associations between disulphide
concentration, disulphide bonding pattern, loop sizes, Structural Classification of
Protein (SCOP) fold, and PDB header.
Having characterised the common features of protein-protein and protein-ligand
interactions, a tool was developed to perform rapid and accurate scaffold searches. This
technique can be applied to any set of vectors, and its performance was tested on all of
the potential protein-protein scaffolds in the PDB.
A production system was implemented featuring performance between 4x to 13x times
better than the state of the art (on some queries the system was 25x faster). With more
computational resources available, we expected a normalised performance improvement
of 100x over the state of the art. A novel scaffold searching technique that combines
the advantages of bitmaps (computationally simple comparisons), hashing-indexing
(fast searches) and distance tolerances (high recall) was developed, however it was not
integrated into the production scaffold-search system as the technique was notcompletely suitable for the problem domain.
This work contributes to a better understanding of a little explored aspect of drug design
and development of paramount importance – that of the privileged structure
counterparts of proteins. This includes understanding at both the macro and micro
level: common amino acid arrangements and the physicochemical features on the
surfaces of proteins. Our results support the theory that privileged structures acquire
their remarkable facility by binding to features conserved across diverse receptors.
To extend and apply this finding, we explored the development of drugs to target
‘undruggable’ protein-protein interactions through the use of privileged protein
scaffolds. We extracted the characteristic attributes of these scaffolds and developed an
efficient tool to enable appropriate scaffolds to be identified that have the ability to
modulate a protein-protein interaction of interest.