Reactivity data for prediction models in organic synthesis

Predictions models for reactivity have been one of the key underpinning tools for synthetic science. They are critically important for predicting reaction selectivity, synthetic route planning/reagent selection and in predicting impurities in the desired products. The development of these models has been difficult, despite recent advances in machine learning and data science, due to the lack of trustworthy kinetic and reactivity data in the open literature. The only rigorous source of reactivity data is Herbert Mayr's database, 1 but its scope is limited.

Thus, there is an urgent needs for: (i) a robust sector-wide reactivity data collection framework, i.e. standardized reactions and measurements, in both chemical industry and academia; (ii) a mechanism to publish and share such data; and (iii) experimental methodologies to generate such data extremely quickly without sacrificing data quality.

Proposed Project: we aim to address the issues highlighted above through an experimentally focused project (in conjunction with on-going machine learning activities in my group toward reactivity prediction). The objectives of the project are:

O1: Adjustments and application of the reactivity data protocol developed at Lhasa to nitrosamines, an important class of genotoxins.

O2: Development of competition reaction framework for reactivity/mechanistic data collection.

O3: Development of very high-throughput kinetic/reactivity data collection platforms based on non-contact spectroscopy and advanced data processing.