Fig. 5: Structural equation regression identifies significant correlations between gene expression and cellular mechanics accounting of spatial confounding effects.
From: A computational pipeline for spatial mechano-transcriptomics
a, A volcano plot for dataset 2 showing, for each gene, the adjusted (adj.) P value by two-sided t-test followed by BH adjustment (y axis) plotted against the regression coefficient, βspatial, obtained by regressing the spatially regressed residual of gene expression on the spatially regressed residual of cellular pressure (x axis). Side plots represent such linear regressions for two example genes whose spatially regressed gene expression residuals are respectively negatively (left) and positively (right) associated with the spatially regressed residual of cellular pressure. b, GO overrepresentation analysis for up- and downregulated genes with cellular pressure. GO terms are ranked according to P value by hypergeometric test and gene count. c, A volcano plot for dataset 2 showing, for each gene, the adjusted P value by two-sided t-test followed by BH adjustment (y axis) plotted against the regression coefficient, βspatial, obtained by regressing the spatially regressed residual of gene expression on the spatially regressed residual of the magnitude of the cellular stress tensor. Side plots represent such linear regressions for two example genes whose spatially regressed gene expression residuals are respectively negatively (left) and positively (right) associated with the spatially regressed residual of the cellular stress tensor. d, GO overrepresentation analysis for up- and downregulated genes with cellular stress tensor. GO terms are ranked according to P value by hypergeometric test and gene count. e, Spatial gene expression maps for selected genes (labeled in purple in a) displaying significant correlations between gene expression and cellular pressure in both the linear and the structural equation regression analyses.
