Fig. 3: Batch correction and label transferring benchmarks.
From: MASI enables fast model-free standardization and integration of single-cell transcriptomics data
a Comparison of label transferring for MASI, supervised, and semi-supervised methods. ACC: overall accuracy. Macro F1 is the average of F1 scores per cell type. A higher score in both metrics suggests better cell-type prediction. b Comparison of batch correction for MASI, scNym10, scArches scANVI11, Symphony17, Scanorama16, LIGER6 and Seurat7. Cell-type silhouette score measures how well the integrated representation by these methods preserves cell-type variation, while the batch entropy mixing score measures how well the same cell type from different batches is mixed. c Visualization of MASI integration through UMAP. Cells are colored according to MASI-reported annotation (top left), author-reported annotation (top right), and batch id (bottom left).
