Anomalous ¹H NMR chemical shift behavior of substituted benzoic acid esters

Abstract

Benzoic acid esters represent key building blocks for many drug discovery and development programs and have been advanced as potent PDE4 inhibitors for inhaled administration for treatment of respiratory diseases. This class of compounds has also been employed in myriad industrial processes and as common food preservatives. Recent work directed toward the synthesis of intermediates for a proprietary medicinal chemistry program led us to observe that the 1H NMR chemical shifts of substituents ortho to the benzoic acid ester moiety defied conventional iterative chemical shift prediction protocols. To explore these unexpected results, we initiated a detailed computational study employing density functional theory (DFT) calculations to better understand the unexpectedly large variance in expected versus experimental NMR chemical shifts.

Rapid prediction of full spin systems using uncertainty-aware machine learning

“…Advances in DFT have allowed it to produce high quality shi and coupling values 3,4 which allows them to be used in a variety of practical applications and as a common point of comparison. 1,5,6 Towards the lower levels of theory are machine learning (ML) techniques, which primarily use data to determine their predictions. We explore the advancements in ML models in Section 2.…”