Fig. 6

Ambiguity in classification prediction probabilities informs whether a sample is from a new origin. a The distributions of Simpson index on the class prediction probabilities of each sample based on 10-fold (red) and leave-one-city-out (blue) cross validation settings, which indicate the diversity pattern for samples from pre-trained or new origins, respectively. b The receiver operating characteristics curve of the Bayes classification model on predicting new-origin status using the Simpson index values computed through a leave-one-out design. c Prediction of new-origin status on mystery samples from new cities