Identifying and developing new drugs is a time consuming and expensive process, and the attrition rate of the drug design process remains high. Using graph-based signatures we have been developing approaches that enable rapid screening of biological activities, along with their pharmacokinetic, pharmacodynamic and toxicity profiles. This can help identify in early stages molecules of interest to focus development efforts upon, reducing failure risks, costs, time and animal usage.
Molecular signatures of biological activity
We are exploring the chemical signatures of molecules that could be used to treat different cancers and microbial pathogens (including TB, malaria and the most dangerous hospital-acquired, antibiotic-resistant infections: Escherichia coli, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa and methicillin-resistant Staphylococcus aureus (MRSA), as well as the common causes of fungal infections Cryptococcus neoformans and Candida albicans).
By mapping the chemical signatures of active and inactive molecules, we are generating novel predictive algorithms capable of identifying potential new antibiotics and chemotherapeutics. In collaboration with our partners, these predicted molecules will be tested in models of the diseases.
We have also successfully used this approach to predict and model a broad range of pharmacokietic and toxicity properties of small molecules. These methods are being implemented as part of medicinal chemistry pipelines in the design of new therapeutics.
small molecules for BIG targets- Targeting protein-protein interactions with fragments
Most proteins work within a network of interactions with other proteins, selectively targeting specific interactions, modulating protein function and providing the opportunity to develop more selective and effective drugs.
But while drugs are usually around 100 Å2, proteins interact tightly using way larger protein-protein interfaces, ranging from 1000–6000 Å2. This raises the challenge of how we can use a small molecule to affect an interface many times larger, which until recently was considered to be flat and undruggable. We and others have had success using fragment-based drug discovery to identify novel protein-protein interaction modulators. This allows us to take advantage of hot-spots within the protein interfaces that mediate a large proportion of the binding energy, growing the molecule to improve binding affinity and drug like properties.
The crystal structures of many protein interface modulators with their targets have been solved, which opens up the possibility for us to ask: what are the major components of binding affinity? and can we use this information to predict fragments likely to bind to a given interface? We are addressing these questions using structural bioinformatics and machine learning, leading to the development of novel programmes, which we validate using biophysical and structural approaches to test fragment binding.