Fig. 1

Overview of bindSC. a Inputs supported by bindSC. BindSC can integrate two single-cell assays such as transcriptomes, epigenomes, spatial transcriptomes, and proteomes. b Bi-order integration of two modalities (X and Y) with unpaired cells and unmatched features using the bi-CCA approach. In the data matrices, each row represents a gene/locus, and each column represents a cell. Step 1: initializing a modality fusion matrix Z linking the two modalities (Methods). Step 2: matching both cells and features across modalities using CCA. Step 3: updating Z using the obtained cell-cell and feature-feature matching results. Steps 2 and 3 are performed iteratively to optimize Z. c Based on canonical correlation vectors (CCVs) in the derived latent space, bindSC can (1) jointly cluster cells in both modalities to define cell types and (2) transfer labels from one modality to another modality. Association of Z and Y measured in the same cell enables to infer gene-protein and peak-gene regulatory networks. d The integrated multiomics feature profiles enable us to (1) link genes to regulatory elements, (2) map RNA expressions to spatial locations, and (3) delineate cells by both RNA and protein signatures