Accurate prognosis for localized prostate cancer through coherent voting networks with multi-omic and clinical data

Abstract

ABSTRACTBackgroundProstate cancer is a very heterogeneous disease, from both a clinical and a biological/biochemical point of view, which makes the task of producing a stratification of patients into risk classes remarkably challenging. In particular, it is important an early detection and discrimination of the more indolent forms of the disease, from the more aggressive ones, requiring closer surveillance and timely treatment decisions.MethodsWe extend a recently developed supervised machine learning (ML) technique, called coherent voting networks (CVN) by incorporating novel model-selection technique to counter model overfitting. The CVN method is then applied to the problem of predicting an accurate prognosis (with a time granularity of 1 year) for patients affected by prostate cancer. The CVN is developed on a discovery cohort of 495 patients from the TCGA-PRAD collection, and validated on several other independent cohorts, comprising a gross total of 744 patients.FindingsWe uncover seven multi-gene fingerprints, each comprising six to seven genes, and a mixed clinical and genomic 5-marker fingerprint, that correspond to different input data types (clinical, mRNA expression, proteomic assays, methylation) and different time points, for the event of post-surgery progression-free survival (PFS) in patients diagnosed with prostate adenocarcinoma, who had not received prior treatment for their disease.With a mixed 5-marker genomic and clinical fingerprint comprising Gleason primary score, tumor stage, psa, and molecular protein expression levels for CDKN1B and NF2 we attain on three independent cohorts statistically significant AUC values of 0.85, 0.88, and 0.87 respectively for PFS prediction at 3 years.For purely genomic fingerprints, in seven independent cohorts for 21 combinations of cohort vs fingerprint, we report Odds Ratios ranging from a minimum of 9.0 and a maximum of 40.0, with average 17.5, geometric mean p-value 0.003; Cohen’s kappa values ranging from a minimum of 0.18 to a maximum of 0.65, with average 0.4; and AUC ranging from a minimum of 0.61 to a maximum of 0.88, with average 0.76, geometric mean p-value 0.001, for PFS prediction at 2, 3, and 4 years.Many of the genes in our fingerprint have recorded prognostic power in some form of cancer, and have been studied for their functional roles in cancer on animal models or cell lines.InterpretationThe development of novel ML techniques tailored to the problem of uncovering effective multi-gene prognostic biomarkers is a promising new line of attack for sharpening our capability to diversify and personalize cancer patient treatments. For the challenging problem of discriminating a fine time-scale for aggressive types of localized prostate cancer, we show that it is possible to attain more accurate prognostic predictions, with a granularity within a year, for the post-surgery early years.