Number of compounds in a respective targeted library chemical space for a given similarity threshold:

Performance of COVID-19 machine learning model in 10-fold cross validation and on external set:

Accuracy in 10-fold cross-validation of the different prediction models for salty and umami taste:

ROC curve for umami and salty taste prediction logistic regression models:

Number of compounds predicted to be (in-)active against SARS-CoV-2 main protease at different confidence score thresholds:

Number of compounds predicted to be sweet/bitter at different confidence score thresholds:

Distribution of the docking scores (binding energy) of the 1078 docked natural compounds:

Summarized results of the 20 natural compounds with the lowest docking score.
(NA = visualization failed, light gray row = natural compound selected for further evaluation):