About CASPRE
CASPRE (CArbon Shift PREdictor) is a hybrid, solvent-aware 13C NMR chemical shift predictor designed to deliver robust accuracy across diverse chemical spaces and solvent environments. It combines 1) a rapid, solvent-aware database search based on HOSE codes (local atom environments) for molecules that have close, experimentally characterized analogs, and 2) solvent-specific graph neural networks based on the FullSSPrUCe architecture, fine-tuned to be solvent-aware and to handle diverse compound scaffolds, including natural products.
To use CASPRE, draw or paste a structure (SMILES, InChI or MOL file) into the MarvinJS applet on the =< link_to 'predictor page', new_carbonpred_path | , choose the solvent (CDCl3, CD3OD or D2O) and the chemical shift reference (TMS, or DSS, TSP or TMS for D2O), and press "Predict". The prediction runs in the background; the result page shows the structure and the predicted 13C shift of every carbon, whether it came from the solvent-matched database or the neural network, and a CSV download of the results.
The data used to train and test CASPRE are available under the =< link_to 'Training Data', training_data_carbonpreds_path | tab.
Citing CASPRE:
- Sajed T, Saha S, Lee BL, Berjanskii M, Sayeeda Z, Wang F, Gautam V, Wishart DS. Accurate, Robust, Solvent-aware Prediction of 13C NMR Chemical Shifts of Organic Compounds Using a Hybrid Machine Learning Approach. Submitted.
Licensing Details
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
This means that our data can be used in derivative works as long as proper attribution is given to the source of the data, but these derivative works may not be used for commercial purposes.