Fig. 1

The workflow diagram of PACSI. A The inputs for PACSI are the scRNA-seq data, the bulk expression data, phenotype labels corresponding to bulk data and PPI networks. B PACSI extracts separately the cell signatures and bulk sample signatures from scRNA-seq data and bulk expression data, respectively (left and right). Meanwhile, PACSI calculated the largest connected component of the PPI network using the R igraph package (middle). C Each cell signature and each bulk sample signature respectively induces a module in the largest connected component of the PPI network, and then the network-based cell-phenotype proximity is calculated (left). Next, the PACSI results are evaluated by calculating empirical P values from random permutations (middle). The PACSI-identified cells can be visualized using UMAP (right). D The cell subpopulations identified by PACSI are used for downstream analysis