CarbNMR
A Web-Based Carbohydrate NMR Spectra Prediction Tool
Abstract
Carbohydrates are structurally complex molecules key to the development of anti-bacterial vaccines. Nuclear Magnetic Resonance (NMR) spectroscopy is a primary experimental technique used to determine carbohydrate structure, but NMR spectra remain difficult to interpret, even for experienced researchers. NextShift, a machine-learning model developed at the University of Cape Town in 2025, uses Graph Neural Networks for NMR spectra prediction. However, this model lacks a user interface and is therefore inaccessible to carbohydrate researchers. Here we present CarbNMR, a web-based carbohydrate NMR spectra prediction tool with a focus on modularity and usability. CarbNMR extends SugarDB, a carbohydrate structure database and visualisation tool, to allow for in-place NMR spectra prediction of known carbohydrate structures. The predicted spectra can be compared against experimental data and exported for further analysis. We developed CarbNMR across three user-centred iterative cycles, each evaluated with carbohydrate researchers using task-based testing and a System Usability Scale questionnaire. CarbNMR achieved a final SUS score of 88, indicating excellent usability. CarbNMR makes accurate NMR prediction available to carbohydrate researchers at the point of structural analysis, improving the accessibility of NMR prediction and spectral interpretation.
Videos 1
Watch presentations, demos, and related content
CarbNMR Video
Gallery 1
Explore the visual story of this exhibit
Title Image