Fig. 3. Metadata of sample’s parent dataset is useful but insufficient to infer sample annotations.

A Scatterplot of the area under the precision-recall curve (auPRC) scores of sample tissue predictions from just sample text (x-axis) vs. combination of predictions from both sample and dataset text (y-axis). B) Scatterplot similar to (and shares y-axis with) panel A but with auPRC scores of predictions from just dataset text on the x-axis. Combination of dataset-and-sample-based predictions versus the dataset-based predictions, where the predicted probability for an experiment is used as the prediction for all samples in that dataset. Each point in the scatterplots correspond to a tissue/cell-type term. auPRC scores are averages across cross-validation folds. The solid line denotes equal performance between the two methods. Source data are provided as a Source Data file.