Fig. 4

Benchmarking the performance of WISE against seven methods in five datasets. A The fraction of accurately predicted clusters, incorrectly predicted clusters, and unassigned clusters in the predictions made by all tools across five datasets. Different cell marker databases were used as input for each tool, including self-built databases and publicly available databases like PanglaoDB, CellMarker 1.0, CellMarker 2.0, and clustermole. The last 12 tissues were from the Tabula Sapiens dataset. The results for Lung_Trachea were not shown as this tissue was absent in many cell marker databases. B Accuracy comparison for all methods