Science Spotlight

Expressing the potential for prostate cancer death

From the Stanford Lab, Public Health Sciences Division.

Prostate cancer is the most common non-skin cancer among American men, and more than 26,000 are expected to die from the disease in 2017.  As most prostate tumors are diagnosed at an early localized stage, identifying those that are more likely to progress to metastatic disease or death could help guide treatment decisions and improve outcomes. Tumor-derived molecular signatures are expected to help with this risk prediction. Dr. Rohina Rubicz, a former Fred Hutch post-doc working in the lab of Dr. Janet Stanford in the Public Health Sciences Division, performed a genome-wide analysis to identify new prognostic biomarkers. In a recent issue of Molecular Oncology, the authors describe a panel of 23 mostly novel transcripts that have clinical potential to improve patient stratification in the future.

Currently, the best clinical variable for determining prostate tumor aggressiveness and metastatic potential is the Gleason score, a grading system based on how a tissue biopsy appears under a microscope. Measuring gene expression within the tumor has the potential to provide complementary discriminatory information, as it can reveal genetic mutations and epigenomic modifications that might influence metastatic progression. Currently available commercial gene expression panels were built to capture information for only a limited number of candidate genes or pathways, suggesting a genome-wide approach may reveal additional informative biomarkers. Said senior author Dr. Stanford, “this project took a different approach (classification as opposed to association) for identifying a panel of differentially expressed mRNAs in primary prostate cancer tumor samples.”

To evaluate gene expression, the authors used previously-collected tumor tissue blocks from 383 men with clinically localized prostate cancer who underwent radical prostatectomy.  Patients were followed over time to identify those that remained disease-free for at least 5 years and those that progressed to metastatic-lethal events. From these samples, the authors generated genome-wide expression data on 26,000 transcripts, which were then ranked according to their ability to classify metastatic-lethal versus non-recurrent cases. Classification ability was evaluated based on the area under the receiver-operating characteristic curve (AUC), which measures overall performance, and the partial AUC (pAUC), which can evaluate performance at a high specificity (in this case, 95%). Said Dr. Stanford, “the pAUC was used because we aimed to find biomarkers with a low false-positive rate for classifying men at higher risk for having an aggressive tumor.” Top-ranked transcripts were then evaluated in combination with the Gleason score, to identify those that contributed additional discriminatory information. Positive transcripts were then validated in a separate dataset of 78 prostate cancer patients treated at Eastern Virginia Medical School.

This genome-wide approach identified 23 gene transcripts that were useful for patient classification. Said Stanford, “the majority of these individual mRNAs improved upon Gleason score for prediction of adverse patient outcomes,” raising the AUC from 0.80 using Gleason score alone to 0.83-0.88. Furthermore, “most were unique as they were not previously included in other gene expression panels for predicting biochemical recurrence or metastatic progression.” Evaluation of the genes represented by these 23 validated transcripts suggested several broad functional categories with diverse biological properties related to tumor aggressiveness (see figure).

If further validated, the biomarkers from this study may help improve the prognostic power of gene expression panels beyond those currently available. Said Stanford, “The next step includes further efforts to validate the mRNAs, combine those that are most predictive of metastatic-lethal events into a gene expression score, and validate the score in independent datasets. We are currently collaborating with GenomeDx Biosciences, Inc. for further evaluation of our biomarker panel for predicting at the time of surgery which patients likely have an aggressive tumor and may need closer surveillance and adjuvant therapy.” This would provide important clinical information to cancer patients and providers, and a successful example of precision medicine in action.

Also contributing to this project from the Fred Hutch were Drs. Jonathan Wright, Ilsa Coleman, Catherine Grasso, Milan Geybels, Amy Leonardson, Suzanne Kolb, Daniel Lin, and Peter Nelson.

Gene pathway map
Expression of 11 of the 23 genes identified in this analysis may be modulated by ligand-dependent nuclear receptors, as determined by an Ingenuity Pathway Analysis (IPA). Image modified from the publication.


Rubicz R, Zhao S, Wright JL, Coleman I, Grasso C, Geybels MS, Leonardson A, Kolb S, April C, Bibikova M, Troyer D, Lance R, Lin DW, Ostrander EA, Nelson PS, Fan JB, Feng Z, Stanford JL. Gene expression panel predicts metastatic-lethal prostate cancer outcomes in men diagnosed with clinically localized prostate cancer. Mol Oncol 2017; 11(2):140-150. doi: 10.1002/1878-0261.12014.


Funding for this study was provided by the NIH, Fred Hutch, and the Prostate Cancer Foundation.