Fig. 1: Workflow of SimiC.
From: SimiC enables the inference of complex gene regulatory dynamics across cell phenotypes
a Example of input data to SimiC consisting of the imputed cell expression data of cells with phenotype 1 (C1) and phenotype 2 (C2). The data is visualized in the tSNE-reduced space, where the color indicates the cell phenotype. b SimiC objective function based on LASSO32 with an added similarity constraint. The W(i)s represent the inferred incident matrices, and the X(i)s and Y(i)s the expression matrices of the driver and target genes, respectively. The superscripts correspond to the cell phenotypes (K in total). λ1 and λ2 are hyperparameters to control the weights of the regularization terms. c Incidence matrices inferred by SimiC, one per phenotype. The dimension is the number of driver genes, such as transcription factors (TFs), by number of target genes. Weight Wi,j represents the influence of the ith TF in the jth target gene. The bar plots show the differences in the inferred weights for a given TF and target gene for the two phenotypes (indicated by color).
