Transferable and Interpretable Prediction of Site-Specific\nDehydrogenation Reaction Rate Constants with NMR Spectra

Accurate\nand efficient determination of site-specific reaction\nrate constants over a wide temperature range remains challenging,\nboth experimentally and theoretically. Taking the dehydrogenation\nreaction as an example, our study addresses this issue by an innovative\ncombination of machine learning techniques and cost-effective NMR\nspectra. Through descriptor screening, we identified a minimal set\nof NMR chemical shifts that can effectively determine reaction rate\nconstants. The constructed model performs exceptionally well on theoretical\ndata sets and demonstrates impressive generalization capabilities,\nextending from small molecules to larger ones. Furthermore, this model\nshows outstanding performance when applied to limited experimental\ndata sets, highlighting its robust applicability and transferability.\nUtilizing the Sure Independence Screening and Sparsifying Operator\n(SISSO) algorithm, we also present an interpretable rate constant-temperature-NMR\n(k-T-NMR) relationship with a mathematical formula. This study reveals\nthe great potential of combining machine learning with easily accessible\nspectroscopic descriptors in the study of reaction kinetics, enabling\nthe rapid determination of reaction rate constants and promoting our\nunderstanding of reactivity.