Data driven approaches to chirality of heteroatoms

Chiral heteroatoms are of key importance in a large proportion of bioactive and synthetic compounds, including natural products, pharmaceuticals, agrochemicals, and catalysts. For the main group elements chirality is commonly encountered in pseudo-tetrahedral compounds that have non-superimposable mirror images. Asymmetric synthesis is currently dominated by accessing carbon stereoselectivity and there has been less innovation for routes to sulfur(IV), phosphorus(III), oxygen, nitrogen and boron heteroatoms.
This project aims to address this gap by developing methodologies capable of delivering a diverse and modular library of enantioenriched compounds with chirality at a heteroatom. Our approaches will harness the stability and transferability of known routes to carbon and sulfur(VI) stereocentres in new areas of chemical space. Rather than relying on traditional optimisation campaigns, which can be laborious, we will exploit data science techniques to expedite the process. For example, a comprehensive mapping of the relevant chemical space will support a multi-objective optimisation framework to obtain general methodologies. Specifically, unsupervised machine learning will be applied for exploration and supervised machine learning to produce predictive models and general rules. Overall, by applying a digital chemistry approach, we will be able to access enantioenriched chiral heteroatoms of high value target molecules in a highly efficient manner.