Fig. 1: The design and performance of SCALEX for single-cell data integration.

a SCALEX models the global structure of single-cell data using a variational autoencoder (VAE) framework. b UMAP embeddings of the PBMC dataset before and after integration by indicated methods. Cells are colored by batch (left) and cell-type (right). Misalignments are highlighted with red circles. c Scatter plot comparing SCALEX and the other state-of-the-art single-cell data integration tools in terms of the ARI score (y-axis) and the NMI score (x-axis), based on the Leiden clustering results in the latent space across the indicated benchmark datasets. d UMAP embeddings of the SCALEX integration of the Human Fetal Atlas dataset after integration by SCALEX, colored by batch and cell-type. e Comparison of computation efficiency based on datasets of different sizes sampled from the whole Human Fetal Atlas dataset) including runtime (left) and memory usage (right). Online iNMF was not successfully tested on 4 M data due to a HDF5 file conversion issue for large data (“Online iNMF and LIGER (LIGER, v1.0.0)” subsection of “Comparison with other integration methods” in Methods). f UMAP embeddings of the mouse brain scATAC-seq dataset before (left) and after integration (middle, right); colored by data batch or cell-types. g UMAP embeddings of the PBMC scRNA-seq and scATAC-seq cross-modality dataset before (left) and after SCALEX integration (middle, right); colored by batch or cell-type.