Evaluation of new prostate cancer specific secretory protein panel in serum of prostate cancer patients Free

2664

The promising information on prostate cancer (CaP) associated consistent gene/protein alterations are beginning to emerge from the cumulative data sets derived from few hundred normal prostate and tumor specimens. We constructed a “high confidence” list of six secretory proteins (SPOCK, F5, CRISP3, PLA2G7, TMEFF2, and NPY) as a CaP diagnostic secretory protein panel. We aimed to evaluate the clinical value of each marker candidate in comparison to prostate specific antigen (PSA), and to develop artificial neural networks (ANNs) with the panel inputs to assess whether multivariate analysis of the secretory protein panel in ANNs can decrease unnecessary prostate biopsies.

The panel was defined by integrated evaluations of epithelial cell transcriptome from 50 matched pairs of laser capture micro-dissected benign and malignant prostate epithelial cells, confirmation in other prostate cancer databases and information available on the proteins of interest in tissue and serum. The levels of panel proteins were measured from the sera of 128 controls and CaP patients using sandwich ELISA. ANNs based on the levels of panel proteins (inputs) were constructed using leave-one-out method and compared to the diagnostic performance with and without clinical information by receiver operating characteristics (ROC) analysis.

SPOCK and PLA2G7 could individually distinguish the CaP from the control group with less or similar power to PSA, while CRISP3, F5, TMEFF2 or NPY couldn’t discriminate the groups with statistical significance. However, CRISP3, F5, and NPY could individually identify different subsets of patient cohorts. The ANNs with various combinations of inputs including PSA and excluding TMEFF2 reached the area under curve (AUC) of 0.86 in the ROC analysis. At the 80% sensitivity, the difference in specificity between the ANN model with PLA2G7/PSA (75%) and PSA alone (49%) was significant without adding any other clinical information, e.g. prostate volume.

The prostate cancer specific secretory protein panel as defined by the integrative prostate cancer genomics, knowledge based protein informatics and targeted experimental approaches reported here have potential to provide highly promising biomarker candidates in serum for more precise prostate cancer diagnosis. Targeted panel approach may also have potential to construct biomarker candidates with improving prostate cancer prognosis.