Molecular Dynamics simulation and Machine-Learned Chemical Shifts Predict NMR Spectra and Dynamics in Amorphous Drug Materials [dataset]

Understanding structure and dynamics in amorphous drugs is crucial to understanding their stability, but typical experimental probes provide limited molecular-level insight. Molecular Dynamics simulations were used to model the amorphous forms of the drug irbesartan. Linewidths of 13C and 15N spectra generated from chemical shifts predicted by the ShiftML2 machine-learning model were in excellent agreement with experiment, demonstrating that the effects of fast motions (“β relaxation”) were correctly described. Measurements of 13C shift anisotropies confirm the results of MD simulation in which the glass transition has no impact on molecular reorientation, while 1H T1ρ measurements indicate that diffusional motion dominates the slower dynamics (“α relaxation”). Together, MD and NMR provide a comprehensive picture of structure and dynamics in these systems either side of the glass transition, with excellent agreement between computational and experimental results.