Fig. 3: Cell communication networks and biomarker analysis results.

Heatmaps showing predicted communication networks for PoPH liver peri-central clusters (a), ordered by total relative strength of gene expression (fold-change in gene expression colored in green) across all clusters on the y-axis, with given clusters on the x-axis. Comparing predicted communication for PoPH peri-central clusters, PoPH peri-portal clusters, and non-PoPH cirrhosis peri-central clusters, signaling through the GDF, VCAM, and SEMA7A pathways were the only predicted pathways unique to PoPH peri-central clusters (b, left). Relative contribution refers to the ratio of how much a given receptor-ligand pair contributes to the full communication network (defined by GDF, VCAM, and SEMA7A signaling). Sending of the signal (secretion of GDF15) and mediation/influencing was most strongly predicted in the peri-central Hepatocyte 9, Cholangiocyte 4, and HSC S1 clusters (b, right). Significantly (Kruskal–Wallis test p < 0.05) lower levels of circulating plasma BMP9 are seen in PoPH samples relative to both non-PoPH cirrhosis and IPAH in the initial biorepository cohort (c), there is no significant difference in plasma BMP9 levels between IPAH and non-PoPH cirrhosis groups. The dashed red line represents the lower limit of detection of the plasma BMP9 assay. As demonstrated by ROC Curves, multivariable regression analysis confirms plasma BMP9 discriminates PoPH from non-PoPH cirrhosis (d) and PoPH from a mixed cohort of non-PoPH cirrhosis and IPAH subjects (e) in the initial biorepository cohort.