Associations among signaling molecules from tumors and livers, and tumor pathophenotypes. (A,B) Multivariate analysis revealed relationships among tumor pathophenotypes and representative molecules from signaling pathways obtained from tumors (A) and liver (B). We show analogous biplots representing only the signaling proteins from tumors and livers without pathophenotype variables elsewhere (Figure S5A,B in Additional file 2). (C) Schematic of the correlations between specific signaling molecules from tumors and livers, together with different tumor traits. All correlations with P < 0.05 were included. The correlation coefficient r is represented by a red-blue scale and for non-significant correlations is represented homogeneously in gray. We show the complete information from this figure in Table S9 in Additional file 1. (D) Correlation between pAKT1 and total AKT1 in liver and tumors. There was a strong correlation between total and phosphorylated levels of AKT1 in liver (P < 0.0001), but not in tumors. As indicated in the manuscript, this means that the pAKT/total AKT ratio in the liver is constant, but not in the tumors, indicating that the percentage of total AKT phosphorylated in the livers of these mice is always the same, while in the tumors it is random. (E) Distribution of some protein signaling molecules from tumors and livers across the different prognosis clusters. Note the similarities between these signaling molecules from tumors (red) and livers (green) regarding differential statistically significant levels among prognosis clusters. We include further, similar examples in Figure S5C in Additional file 2. The values represented in (D) are provided in Table S10 in Additional file 1.