Skip to main content
Figure 6 | Proteome Science

Figure 6

From: Automated quantitative multiplex immunofluorescence in situ imaging identifies phospho-S6 and phospho-PRAS40 as predictive protein biomarkers for prostate cancer lethality

Figure 6

Outline of statistical analysis flow. For each patient, two tissue cores from the highest Gleason area were placed into TMA blocks. Mean values of biomarker expression in the tumor epithelium region of each TMA core were used for analysis, resulting in two biomarker values per patient. For PTEN, SMAD4, and pS6, the lowest value from the two cores was used for analysis. For CCND1, SPP1, p90RSK, pPRAS40, and FOXO3, the highest value from the two cores was used. Using these values, 10,000 bootstrap training samples were generated and both multivariate Cox and Logistic Regression models were trained on each training sample. Testing was performed on the complement set. Given the cohort included censored data, we used both CI and AUC to estimate the model performance. The marker combinations that were tested in the models were as follows: 4 markers (PTEN, CCND1, SMAD4, SPP1), 3 markers (CCND1, SMAD4, SPP1), and 3 markers with each of the following combinations of phospho markers: pS6, pPRAS40, and [pS6 + pPRAS40].

Back to article page