Table 1.

Prediction of pancreatic cells. The training panel corresponds to the Muraro, Segerstolpe, and Xin datasets used as a reference to train the prediction models for each cell type from the islets of Langerhans. As part of the training, no other cell types were considered. The test information corresponds to the Baron [31] dataset used to measure the performance of the trained models in an independent dataset. The Baron dataset contains epsilon, acinar, stellate, ductal, endothelial, Schwann, and T cells referred as “Other” in this table. The accuracy is defined as the fraction of cells correctly assigned for each cell type of interest. The accuracy for the remaining cells corresponds to the fraction of cells from the test dataset that are correctly unassigned to any of the classes of interest (negative controls) as a consensus across all four prediction models