Welcome!

PrediSweet lets you predict the sweetness of a molecule based on its structure.


Each molecule is first converted to a set of descriptors.
Then a machine-learning model predicts the sweetness (in logarithmic scale) based on these molecular features.
Finally, the quality of the prediction is assessed based on 3 metrics.
More details can be found in the Help section

If you have used PrediSweet in a scientific publication, please cite the following paper:
C. Bouysset, C. Belloir, S. Antonczak, L. Briand, S. Fiorucci, Novel scaffold of natural compound eliciting sweet taste revealed by machine learning. Food Chem. (2020). DOI: 10.1016/j.foodchem.2020.126864

For testing purposes, you can submit jobs without an account by following this link, but the results could be accessed by anyone.
Consider logging in or signing up to get a private access to all of your jobs and molecules.

Internet Explorer and Microsoft Edge are not fully supported browsers. Please use Firefox or Chrome instead.