NIR-II emission wavelength prediction
This web app serves a retrained mainline model for solvent-aware NIR-II molecular emission wavelength prediction. The workflow builds Morgan fingerprints, RDKit fragment counts, and solvent one-hot features, then applies training-set feature screening to obtain the fixed feature table used at deployment.
After users train the model on their own data, the result panel reports the base XGBoost prediction. If an OOF nearest-neighbor residual library is supplied, the optional correction term and neighbor-similarity evidence are also shown.
Input
One molecule SMILES and one normalized solvent label.
Mainline
Tuned XGBoost trained on the feature set selected from the training data.
Output
Base prediction, optional OOF-NN correction, similarity evidence, and final wavelength.