Fig. 2
From: scAlign: a tool for alignment, integration, and rare cell identification from scRNA-seq data
scAlign outperforms existing alignment approaches on four benchmarks. a CellBench, a benchmark consisting of mixtures (mt) of RNA from three cancer cell lines sequenced using multiple protocols. Plots from left to right: (1) UMAP plot of embeddings after alignment with scAlign, where each point represents a cell, and cells are colored according to their mixture type (mt) as reported in Tian et al. (2) UMAP plot of embeddings after alignment with supervised scAlign (scAlign+). (3) Bar plot indicating the accuracycomposite (see the “Methods” section) of a classifier, measured as a weighted combination of cross-condition label prediction accuracy and alignment score. b Same as a, but with the Kowalczyk et al. benchmark consisting of hematopoietic cells sequenced from young and old mice. Cells are colored according to type (LT, ST, MPP, legend at bottom). c Same as a, but with the Mann et al. benchmark consisting of hematopoietic cells sequenced from young and old mice, challenged with LPS. d Same as a, but with the HeterogeneousBenchmark dataset consisting of hematopoietic cells responding to different stimuli