CatSci Sponsor LabLinks Event – Application of Digital Tools to Synthesis
On 17th June 2021, CatSci sponsored a LabLinks event on the Application of Digital Tools to Synthesis. The online event consisted of a trio of exciting presentations led by industry experts, including Dr Connor Taylor and Kobi Felton, both from the University of Cambridge, as well as SAS’s Ben Francis.
To begin the virtual symposium, Dr Connor Taylor delivered a presentation on Modern Kinetic Analysis. This talk showcased Dr Taylor’s Compuentics application, which allows chemists who have little recollection of kinetics training to build a better understanding of the process. Whilst the subject has moved on from attempting to transfer concentration and time data so a straight line can be drawn, many of the applications that operate in this area come with a substantial cost barrier. Compunetics is a low cost solution that makes kinetic fitting easy for chemists armed with possible mechanisms that could be operating in their reaction, and time course data on how the concentrations of the participating species change. Connor’s presentation offered an example of how this novel application can be used to build a better understanding around an NHC-catalysed electrochemical reaction. With a working kinetic expression to hand, Connor then described how it can be used to simulate processing scenarios in lieu of wet experimentation. The presentation concluded with a recent update of the software, which uses a super-modified simplex algorithm to guide a chemist who wants a more efficient means of optimising a process, rather than changing one factor at a time. Compunetics has an intuitive interface and its use does not require any coding experience.
Next up on the agenda, Kobi Felton presented a talk entitled “Benchmarking machine learning methods for reaction optimisation,” and explained his creation, Lably. This is an application that Kobi has produced for reaction optimisation using machine learning techniques. Crucially, Lably offers confidence that an appropriate machine learning method is applied to a problem using highly scalable simulations of reactions called benchmarks. Firstly, Kobi described their use with an SNAr example where a trained benchmark was used to compare seven optimisation approaches. A second example concerned the optimisation of the categorical and continuous variables associated with a C-N cross-coupling reaction. Over just 20 reactions, the reaction yield was raised to ca. 95%. Kobi closed his talk with a call to interested parties interested in further development of the tool.
Ben Francis rounded off the event with his talk on “Optimising reactions when the answers aren’t in your data.” This final presentation was a succinct story on the analysis of big datasets, as well as the challenge of systematically studying problems that are described by categorical variables.
The virtual event was insightful, educational, and informative. All attendees left feeling inspired and having learned something new. As well as the presentations, participants were invited to network with the speakers and other attendees in the LabLinks LinkUp lounge. It was clear from the meeting that for synthetic chemists to operate in today’s environment, a little bit of understanding of data science can go a long way, and there are many opportunities out there to enhance our knowledge and understanding.