Fig. 4: Group comparisons and biomarker analysis on the training cohort.

A Combined Up Set [94] and volcano plots, with color-coded annotations of diverse RPPA endpoint sets that characterize specific binary comparisons of sample groups. The vertical histogram (upper panel) reports the relative frequency of unique (single dots) and shared (connected dots) significant RPPA antibodies resulting from three specific sets of statistical comparisons, namely PCa versus Hypertrofic, HD and Post-prostatectomy disease free (DF), respectively. The volcano plot in the lower panel shows fold-changes (log2) versus significance [−log10(p - value)] for all analyzed comparison sets (i.e. PCa versus Hypertrofic or HD or disease-free, respectively). The color-coding (‘queries’) of RPPA antibody labels in the volcano plot matches the corresponding colors sets in the main frequency histogram. The horizontal histogram adjacent to comparison sets (left-bottom part of the upper panel) shows the absolute frequency of statistically significant RPPA antibodies obtained for each individual set. B Univariate ROC curve analysis of two selected, significant candidate markers resulting from statistical comparison of controls (grouped HD plus Hypertrophic diseases) and tumors (upper plots). The diagnostic performance of combined top-scoring candidates was further assessed by ROC curve analysis (bottom plot). All plots report the AUC value along with the 95% confidence interval as well as p, optimal cut-off, sensitivity and specificity values. C Bivariate plot of best candidates (c-Myc T58/S62 and TSC2 Y1571) normalized RPPA intensity values in the analyzed cohort comprising tumors (PCa) and control samples, i.e. HD and Hypertrophic diseases, the latter subdivided into Adenoma and Inflammatory diseases, being the category ‘others’ referred to grouped non-neoplastic samples (controls). The ellipses represent the probability of distribution (95% confidence under normality assumption) for the two main comparison groups, namely PCa and others.