3.1 Introduction to General Design Considerations
57
Fragment-based design can also be initiated by accessing starting fragments in
the small to big approach, which can be a powerful one to develop new medicinal agents and has been used to develop antibacterial agents. In a neat example
of this, a fragment-based approach combined with X-ray crystallography as a
primary screen identified four hit compounds (from 352 fragment compounds)
targeting the Escherichia coli bacterial sliding clamp, a protein vitally important
in DNA replication. A tetrahydrocarbazole derivative then emerged as a good lead
compound from further work and the methyl- and ethyl ester derivatives of this
compound showed moderate antibacterial activity against some Gram-positive and
Gram-negative bacteria (Yin et al. 2014). Further development of this intereting
work has identified binding sites on the sliding clamp for tetrahydrocarbazole-based
compounds (Yin et al. 2015).
In silico-based design
Various in silico approaches have been used to help in the design of multi-targeting
ligands including those aimed at antibacterial activity (Ma and Chen 2012). Machine
learning techiques have shown considerable potential in identifying new antibacterials (Ivanenkov et al. 2019b). Ivanenkov and co-workers started with a very large
compound library (140,000 small compounds) with antibacterial activity against
Escherichia coli and activity being assessed under the same conditions in the one
assay. The compunds had a range of structures outside those of previously described
antibacterials. Mining of this data then resulted ultimately in the identification of
several compounds with potent activity in vitro and in vivo, although structural
details were not revealed on these and very little on the mode of action apart from
a mention of translation inhibition in prokaryotes and the induction response (to
DNA damage) in some instances. Any multi-targeting mechanisms of action, while
possible, await further assessment.
An in silico strategy was also employed in laying the basis for new drug designs
for the treatment of the bacterial disease leprosy caused by Mycobacterium leprae
or the more recently identified Mycobacterium lepromatosis. In this work the MUR
enzymes (UDP-N-acetylmuramic acid or muramoyl containing ligases MurC, D,
E and F) involved in the peptidoglycan pathway were targeted. Conserved or classspecific amino acid residues were identified, which then pointed to key interaction
points to be taken into account in the design of any single molecule multi-action
drugs (Anusuya and Natarajan 2012).
The powerful use of artificial intelligence (AI) and other computer-based methods
for analysing multi-targeting in the context of antibacterial activity is presented in
recent work by Abrusán and Marsh (2019). They based their work on ligands binding
to sites across multiple protein chains. These sites were considered as likely to be
conserved ones and thus suitable for targeting by broad-spectrum antibacterials. The
multi-site binding ligands, which arose from the de novo ligand design plus deep
learning, often contained, for example, sub-structural fragments with a degree of
likeness to those in known antibacterials. Intriguingly, although perhaps understandably, this work revealed that multi-site binding also favoured compounds that were
57
Fragment-based design can also be initiated by accessing starting fragments in
the small to big approach, which can be a powerful one to develop new medicinal agents and has been used to develop antibacterial agents. In a neat example
of this, a fragment-based approach combined with X-ray crystallography as a
primary screen identified four hit compounds (from 352 fragment compounds)
targeting the Escherichia coli bacterial sliding clamp, a protein vitally important
in DNA replication. A tetrahydrocarbazole derivative then emerged as a good lead
compound from further work and the methyl- and ethyl ester derivatives of this
compound showed moderate antibacterial activity against some Gram-positive and
Gram-negative bacteria (Yin et al. 2014). Further development of this intereting
work has identified binding sites on the sliding clamp for tetrahydrocarbazole-based
compounds (Yin et al. 2015).
In silico-based design
Various in silico approaches have been used to help in the design of multi-targeting
ligands including those aimed at antibacterial activity (Ma and Chen 2012). Machine
learning techiques have shown considerable potential in identifying new antibacterials (Ivanenkov et al. 2019b). Ivanenkov and co-workers started with a very large
compound library (140,000 small compounds) with antibacterial activity against
Escherichia coli and activity being assessed under the same conditions in the one
assay. The compunds had a range of structures outside those of previously described
antibacterials. Mining of this data then resulted ultimately in the identification of
several compounds with potent activity in vitro and in vivo, although structural
details were not revealed on these and very little on the mode of action apart from
a mention of translation inhibition in prokaryotes and the induction response (to
DNA damage) in some instances. Any multi-targeting mechanisms of action, while
possible, await further assessment.
An in silico strategy was also employed in laying the basis for new drug designs
for the treatment of the bacterial disease leprosy caused by Mycobacterium leprae
or the more recently identified Mycobacterium lepromatosis. In this work the MUR
enzymes (UDP-N-acetylmuramic acid or muramoyl containing ligases MurC, D,
E and F) involved in the peptidoglycan pathway were targeted. Conserved or classspecific amino acid residues were identified, which then pointed to key interaction
points to be taken into account in the design of any single molecule multi-action
drugs (Anusuya and Natarajan 2012).
The powerful use of artificial intelligence (AI) and other computer-based methods
for analysing multi-targeting in the context of antibacterial activity is presented in
recent work by Abrusán and Marsh (2019). They based their work on ligands binding
to sites across multiple protein chains. These sites were considered as likely to be
conserved ones and thus suitable for targeting by broad-spectrum antibacterials. The
multi-site binding ligands, which arose from the de novo ligand design plus deep
learning, often contained, for example, sub-structural fragments with a degree of
likeness to those in known antibacterials. Intriguingly, although perhaps understandably, this work revealed that multi-site binding also favoured compounds that were