Optimizing Molecular Signatures For Predicting Prostate Cancer Recurrence

Featured Article
Main Category: Prostate / Prostate Cancer
Also Included In: Urology / Nephrology;  Cancer / Oncology
Article Date: 14 Jun 2009 - 0:00 PDT

email to a friend   printer friendly   view / write opinions   rate article

UroToday.com - The mortality rate for prostate cancer is declining due to improvements in earlier detection and in local therapy strategies, however, the ability to predict the metastatic behavior of a patient's cancer, as well as to detect and eradicate disease recurrence remains some of the greatest clinical challenges in oncology.

It is estimated that 25-40% of men undergoing radical prostatectomy will have disease relapse, often termed a biochemical recurrence as the first clinical indication a rising serum level of prostate specific antigen (PSA). The accurate identification of patients at risk for relapse would greatly facilitate the rational application of adjuvant treatment strategies.

The advent of microarray gene expression technology has greatly enabled the search for predictive disease biomarkers. Numerous exploratory studies have demonstrated the potential value of gene expression signatures in assessing the risk of post-surgical disease recurrence beyond the current clinical systems. However, existing molecular predictive models were derived using relatively simple computational algorithms, and the critical issue of whether proposed gene signatures are ready for randomized, prospective clinical validation trials is still under debate in the oncology community. Key to resolving this issue is the development of advanced algorithms that are capable of identifying relevant genes (features in bioinformatic terms) in a background of tens of thousands of genes, and on the basis of a limited number of patient tissue samples. This process is known as feature selection, and achieving this in high-dimensional data remains a major challenge in bioinformatics and machine learning. In order to overcome current restraints, we have derived a feature selection algorithm that addresses several major issues with prior work including computational efficiency and solution accuracy. We have experimentally demonstrated that our algorithm is capable of handling problems with extremely large input data dimensionality, to a point far beyond that needed for gene expression data analysis of genetically complex organisms.

In the study published in The Prostate journal, we conducted a computational analysis to investigate whether the application of our computational algorithm can lead to the derivation of more accurate prognostic molecular signatures for predicting prostate cancer recurrence. To this end, we used a rigorous experimental protocol to compare the prognostic performance of newly identified genetic signatures with those previously derived. Receiver operator characteristic (ROC) curves and survival data analyses demonstrate the superior performance of the new gene signature over previous work. We further derived a hybrid prognostic signature, obtained by integrating gene expression data and clinical variables, that significantly outperformed both the gene signature and the predictive nomogram.

Our results demonstrate that advanced computational modeling can significantly improve the accuracy of molecular prognostic signatures for prostate cancer.

Written by Steve Goodison, MD as part of Beyond the Abstract on UroToday.com

UroToday - the only urology website with original content written by global urology key opinion leaders actively engaged in clinical practice. To access the latest urology news releases from UroToday, go to: www.urotoday.com

Copyright © 2009 - UroToday
Copyright: Medical News Today
Not to be reproduced without permission of Medical News Today