Random forest-based modelling to detect biomarkers for prostate cancer progression

The clinical course of prostate cancer (PCa) is highly variable, demanding an individualized approach to therapy and robust prognostic markers for treatment decisions. We present a random forest-based classification model to predict aggressive behaviour of PCa. DNA methylation changes between PCa cases with good or poor prognosis (discovery cohort with n=70) were used as input. The model was validated with data from two large independent PCa cohorts from the \"International Cancer Genome Consortium\" (ICGC) and \"The Cancer Genome Atlas\" (TCGA). Ranking of cancer progression-related DNA methylation changes allowed selection of candidate genes for additional validation by immunohistochemistry. We identified loss of ZIC2 protein expression, mediated by alterations in DNA methylation, as a promising novel prognostic biomarker for PCa in >12,000 tissue micro-array tumors. The prognostic value of ZIC2 proved to be independent from established clinico-pathological variables including Gleason grade, tumor stage, nodal stage and PSA. In summary, we have developed a PCa classification model, which either directly or via expression analyses of the identified top ranked candidate genes might help in decision making related to the treatment of prostate cancer patients.