DU8ML, a fast and accurate machine
learning-augmented density functional
theory (DFT) method for computing nuclear magnetic resonance (NMR)
spectra, proved effective for high-throughput revision of misassigned
natural products. In this paper, we disclose another important aspect
of its application: correction of unusual reaction mechanisms originally
proposed because of incorrect product structures.