An automated platform for tailored late-stage halogenation of pharmaceuticals

The diversification of lead compounds remains a crucial step in drug development, with bioactive molecules often requiring fine tuning of a range of important pharmaceutical properties. This is achieved through the synthesis of libraries of analogues with subtle modifications in functionality or structure. The complexity of many bioactive molecules, however, does not make this a trivial endeavour. Indeed, many transformations using established chemistry are not suitable due to functional group sensitivity or regiochemical challenges and can require large amounts of time and resources to be spent on multistep syntheses. The development of functional group tolerant transformations can allow the curation of a library of analogues directly from a common, complex precursors through late-stage modification. Harnessing the reactivity of C–H bonds on arenes, which are extremely prevalent in drug like compounds, allows for the addition of functionality to tune electronic and steric properties without the modification of existing functional groups that may be integral to binding properties. Particularly attractive is the installation of a halogen atom to a lead compound to provide opportunities for further diversification through emerging and established cross-coupling techniques as well as the modulation of pharmaceutical properties. Numerous conditions are available to perform C–H halogenation reactions on aromatic compounds. However, “one-size-fits-all” conditions are not reflective of the structural diversity across pharmaceutically relevant molecules and cannot accommodate the density and breadth of functionality. High-throughput screening has been used to optimise C–H functionalisation reactions. However, no such approach exists for the halogenation of pharmaceutical compounds. Here, we develop an automated, machine-learning aided, high-throughput workflow to investigate how varying acidity affects the C(sp 2 )–H chlorination, bromination, and iodination of 32 complex pharmaceutical substrates. Different compounds are shown to require significantly different acidities and yields are improved by optimisation of TFA equivalents. 56 halogenated analogues of 22 pharmaceuticals were prepared using conditions determined by the high-throughput workflow with significant benefit to yield. xTB atomic descriptors were used to predict regiochemical outcomes in a computationally inexpensive way. We envision that the workflow could be used for future similar investigations.