Stanford scientists have developed a machine learning framework that predicts the complete molecular structure of an unknown compound directly from routine 1D NMR spectra with no additional information required, automating one of the most time-consuming and expertise-dependent