Efficient modelling and validation of cryptic protein binding sites for drug discovery

Over 75% of disease-involved proteins cannot be readily targeted by conventional chemical biology approaches. New approaches are needed to increase the scope of molecular medicine.
Cryptic binding pockets, i.e. pockets that transiently form in a folded protein, but are not apparent in the crystal structure of the unliganded apo-form, offer outstanding opportunities to target proteins otherwise deemed 'undruggable' and are thus of considerable interest in academia and the pharmaceutical industry. Unfortunately, not only they are notoriously difficult to identify, but also the molecular mechanism by which they form is still debated. The aim of this collaborative project is to address the knowledge gaps and develop an efficient computational platform based on atomistic molecular simulations to systematically detect druggable cryptic pockets in targets of biopharmaceutical interest. The platform will build on our successful experience in developing and applying enhanced-sampling simulation algorithms to molecular recognition, and will be extensively tested on validated drug targets harbouring cryptic sites. The computational results will be further validated on novel targets by a combination of experiments in collaboration with an industrial partner (UCB).

Drug discovery is a costly and risky undertaking, and the attrition rate of new leads and drug candidates is extremely high and worsening, leading to a significant escalation of drug-development costs (Scannell et al. Nat Rev Drug Discov 2012, 11, 191.). The average costs to develop a new drug increased from the $802 million in 2003 to $2.6 billion in 2014. This leads to lengthy delays in developing new treatments for complex conditions such as cancer, autoimmune and neurodegenerative diseases and huge costs for pharmaceutical companies and the national health service. This proposal will develop and demonstrate a new computational platform addressing a fundamental problem in rational drug design: targeting difficult ("undruggable") disease-involved biomolecules. In so doing it will contribute to greatly expand the number of tractable targets for small molecule therapeutic intervention but also provide a valid alternative to classical substrate-competitive drug design strategies. Our project will lead not only to the development of a unified, user-friendly suite of computational tools for this challenging problem but also to a better understanding of molecular recognition, allosterism and the role of flexibility in ligand-target binding. The approach we propose has the potential to significantly change the rational drug discovery pipeline increasing the predictive value of computational models, the accuracy of free energy calculations and extending their applicability of very complex systems well beyond the capability of current tools. This in turn will speed up the design of more effective and less toxic drugs.
Accordingly, the proposed work has the potential for impact on public healthcare costs and on the speed at which optimal treatments for complex conditions and drug-resistant microorganisms (AMR) can be developed and become available. In addition, the knowledge acquired on allosteric regulation will address a major open issue in structural biology and biophysics, while the platform developed will also be applicable in other fields where flexible molecular recognition is of importance, such as in self-assembled bio-inspired material and industrial biotechnology. In developing accurate chemical models for complex biological systems, this proposal addresses the core questions of the EPSRC CBBC priority area. The novel, ambitious and interdisciplinary approach that we propose will help drug discovery and the EPSRC "Developing Future Therapies" challenge in several ways: 1) the accurate prediction of drug binding mechanisms to cryptic binding sites and the associated binding free energies will enable the early selection of the best candidates in terms of safety and effectiveness; 2) the in-depth understanding of cryptic pocket formation will facilitate the rational design of allosteric inhibitors targeting "undruggable" targets and/or increasing the selectivity (thus decreasing toxicity).
The impact of our novel approach will include the prediction and increased understanding of allosteric small-molecules binding to key biomedical targets in the short term, leading to a rationale for the development of allosteric modulators in the medium term, and thereby to faster and more cost-effective development of new drug treatments (as well as optimized biotechnological production of fine chemicals) in the long term. To achieve maximum impact, a wide uptake of the developed methods and models is imperative. We therefore will collaborate with UCB Slough research center (see support letter) to ensure the wide applicability of the platform and to validate it on pre-competitive (open) drug targets. We plan to make the computational platform freely available, and to spend significant time and effort to make it user-friendly (disseminate the models, methods and tools, as well as the outcomes of the initial applications, through a range of activities and actions, outlined in the specific objectives listed in the PtI.