Conformational landscapes of alkali metalated N-acetylhexosamines: insights from machine learning-assisted ab initio methods and validated by cryogenic infrared photodissociation spectroscopy.
The complex structural effects of metal ion binding in glycans remain a topic of significant scientific interest. In this study, we implement a machine learning-driven workflow utilizing SchNet-based neural network potentials (NNPs) to efficiently map the conformational space of neutral and metalated N-acetylhexosamines (HexNAc) with near-first-principles accuracy. By constructing an extensive database of local minima for all 64 HexNAc isomers, we identify critical trends in how alkali metal ions modulate carbohydrate conformation. Our NNP-driven structural search scheme, validated against cryogenic infrared photodissociation (IRPD) spectra, successfully identifies the specific low-energy conformers responsible for experimentally observed vibrational signatures. Notably, our results reveal a diverse set of structural and energetic trends that vary significantly across the different metalated HexNAc systems. These findings underscore the necessity of explicit, automated structure searching, as the non-intuitive coordination environments of metal ions preclude the use of generalized conformational rules.