Metadynamics and Raman Spectroscopy for Glycan Structure–Spectrum Mapping

Raman spectroscopy offers label-free access to molecular vibrations across biological systems, from small molecules to flexible biomolecular oligomers and whole cells. Yet, understanding how conformational ensembles encode vibrational spectra remains challenging. While small-molecule Raman spectra often permit direct structural interpretation, intermediate-sized biomolecules such as glycans, short nucleic acids, and peptides exhibit compositional and conformational heterogeneity that obscures structure–spectrum relationships. We introduce DynaSpec, an ensemble-resolved structure–spectrum mapping framework that links equilibrium conformational populations to vibrational observables. DynaSpec integrates metadynamics in a machine-learned latent space, population-weighted DFT Raman calculations, and experimental spectroscopy to interpret congested Raman signatures mechanistically. By connecting normal modes to the internal structural coordinates that modulate them, DynaSpec enables bidirectional structure–spectrum inference. Using glycans as a stringent test case, ensemble-averaged DFT predictions show strong agreement with experimental Raman spectra and support >85% classification accuracy across 13 N -glycans. The analysis reveals a conserved α1,6 open–curled switching motif across N -glycans and identifies a glycosidic torsion-enriched spectral window that enables experimental discrimination of positional isomers with >70% accuracy. Applied to binary glycan mixtures spanning structurally distinct pairs to positional isomers, DynaSpec-guided spectral windows enable non-negative least-squares unmixing with R 2 > 0.99 for fucosylation and galactosylation differences, while benchmarking near-superimposable positional isomers. DynaSpec further generalizes computationally to O -glycan cores and glycosaminoglycans, with experimental validation against hyaluronic acid. Together, these capabilities establish an ensemble-resolved framework for structure-informed spectroscopy across diverse biomolecular classes.