Chillin in Flatland

点击翻译按钮获取中文摘要

Innovation in the field of small molecule therapeutics is becoming increasingly difficult. The 'low hanging fruit' of well characterised, easy to drug targets have almost all been picked, yet the need to quickly develop potent new medicines is still a pressing issue. The recent Coronavirus pandemic, and the developing anti-microbial resistance crisis are two obvious examples of this need. Computational methods are emerging as impactful tools in medicinal chemistry and drug discovery, with the ability to screen and analyse vast libraries of compounds far faster than the laboratory scientist. This project aims to develop an open-source, computational tool that allows the investigation of heterocyclic isosteres to discover new cores and scaffolds for compound optimisation and the generation of novel chemical species. It has been proposed that 3D molecules stand a better chance of becoming successful drugs, however over half of the carbons in these '3D' molecules are sp2 hybridised, and therefore flat. Furthermore the synthetic chemistry required to selectively introduce each sp3 carbon is slow, and limited in its scope. Heteroaromatic systems are widely used, and their synthesis well understood in lead optimisation and drug discovery. Their physicochemical properties are more easily tuned, and they are often similar to cellular metabolites and signalling molecules. In 2009 Pitt et al. published VEHICLe, a database of 25 000 potentially accessible small heterocycles of which only 1100 have been reportedly synthesised. We propose to develop a computational tool (the HeteroCycle Isostere Explorer) to discover new heterocyclic cores for compound optimisation by comparing the shape and electrostatic potentials of input molecules to those of each member of the VEHICLe database, and returning the best bioisosteric replacements. This will allow access to new drug-like molecules quickly, but can also be used to determine underexplored areas of sp2 space, which could yield novel physicochemical properties in therapeutic lead compounds. This project would fall within the EPSRC chemical biology research area, and the software development within the computational chemistry area. The aim is to use the tool to generate molecules which would then be synthesised and assayed to test and refine its predictions, and this would then fall also within the synthetic organic chemistry and synthetic biology research areas. When sufficient experimental results have been accumulated, the goal is to develop and incorporate into the tool machine learning models, which can use experimental evidence to guide the predictions. This would then also fall within the AI for Science research area and theme.