Journal Article Page
European UrologyVolume 57, issue 3, pages 363-550, March 2010
Novel Use of a Combined Artificial Intelligence Approach to Identify Patients with Noninvasive Urothelial Cell Carcinoma of the Urinary Bladder Who Are at Greatest Risk for Progression to Muscle-Invasive Disease: A Step Forward
Published online 24 November 2009, pages 407 - 408
Refers to article:
The Application of Artificial Intelligence to Microarray Data: Identification of a Novel Gene Signature to Identify Bladder Cancer Progression
Accepted 27 October 2009
March 2010 (Vol. 57, Issue 3, pages 398 - 406)
The potential for superficial or noninvasive urothelial cell carcinoma (UCC) of the urinary bladder to progress to muscle-invasive disease is quite variable, with progression reported to occur in <1% to >50% of patients  and . Although understaging plays a role, the large amount of biologic heterogeneity in noninvasive UCC of the bladder is also a significant contributor to the large spectrum of UCC patients who are at risk for progression. Identifying patients who are at the greatest risk for progression to muscle-invasive disease would aid in determining the most appropriate treatment for patients with superficial disease , , , and . Therefore, an important focus of bladder cancer research is to identify biochemical and histologic predictors of progression to muscle-invasive disease.
One potential way to identify alterations in gene expression that are correlated with the likelihood of disease progression is to perform microarray analysis on tumor tissue and to attempt to associate changes in the expression of specific genes with individual patient outcomes. When this is done with groups of genes, specific gene “signatures” can be revealed that correlate with a given outcome. This procedure can be used to identify genes that potentially play important roles in the basic biology of disease, and it may also reveal novel therapeutic targets and histologic markers that predict disease progression. However, the cost of microarrays can be prohibitive for use with large numbers of patients and multiple institutes, preventing this approach from becoming a routine clinical test.
In this issue, Catto et al  report their combined use of two forms of artificial intelligence (AI), neurofuzzy modeling (NFM) and artificial neural networks (ANN), to create a “committee of models” to improve the analysis of microarray data, resulting in the identification of a novel gene signature to predict progression to muscle-invasive disease. First, Catto et al ranked the top 200 progression-associated genes identified by their group in a previous study of UCC of the bladder . They accomplished this by using ANN and NFM in two different analytic structures, enabling the researchers to analyze all 200 progression-associated genes simultaneously. Taking the average ranking for each gene into account resulted in the production of a committee-of-models ranking, representing a potential progression-associated gene expression signature for UCC of the bladder consisting of 11 genes. In a subsequent analysis of the original tumor cohort, this gene signature was superior at predicting which tumors would progress compared to conventional statistical analysis as well as to the clinical parameters of stage and grade.
To determine the applicability of this committee-of-models gene signature to predicting progression, Catto et al undertook immunohistochemical analysis using available commercial antibodies on a tissue microarray developed from a separate tumor cohort at another institution. Analysis of staining from different stages of UCC of the bladder revealed a correlation between the tumor pathology and immunohistochemical staining of the identified proteins. More important, the greater number of positively stained signature proteins found in a given tumor resulted in a significant increase in the likelihood of progression to muscle-invasive UCC of the bladder. In fact, immunostaining for members of the committee of models was better for stratifying progression than tumor stage and grade and/or the presence of carcinoma in situ and the multifocality of the cancer.
Catto et al's work is interesting for several reasons. First, they have combined two different AI approaches (NFM and ANN) to identify a gene expression signature that may help stratify patients in terms of their relative risk of progressing to muscle-invasive UCC of the bladder. Following biopsy, immunostaining for these markers may one day help clinicians determine the best course of treatment. Second, the potential future use of novel AI methodologies to interpret the vast amounts of publicly available microarray data is exciting because the methodologies may reveal histologic markers in addition to the 11 identified in this report. Finally, this work has identified several genes that were not previously associated with the progression of UCC. Although these genes were predictive of progression when used only as a group for tissue microarray analysis, it is possible that individual genes may be involved in UCC progression.
In summary, the power of the markers that Catto et al identified for predicting progression in patients is impressive, and hopefully, the markers can be used to identify patients with superficial disease who are truly at risk for progression. The key now will be to determine the suitability of these markers for widespread use to aid clinicians in predicting the potential for progression to muscle-invasive cancer of the bladder.
Conflicts of interest
The author has nothing to disclose.
-  B.W.G. van Rhijn, M. Burger, Y. Lotan, et al. Recurrence and progression of disease in non–muscle-invasive bladder cancer: from epidemiology to treatment strategy. Eur Urol. 2009;56:430-442 Abstract, Full-text, PDF, Crossref.
-  R.J. Sylvester, A.P.M. van der Meijden, W. Oosterlinck, et al. Predicting recurrence and progression in individual patients with stage Ta T1 bladder cancer using EORTC risk tables: a combined analysis of 2596 patients from seven EORTC trials. Eur Urol. 2006;49:466-477 discussion 475–7 Abstract, Full-text, PDF, Crossref.
-  S.S. Chang, M.S. Cookson. Non-muscle-invasive bladder cancer: the role of radical cystectomy. Urology. 2005;66:917-922
-  S.S. Chang, M.S. Cookson. Radical cystectomy for bladder cancer: the case for early intervention. Urol Clin North Am. 2005;32:147-155 Crossref.
-  M.C. Hall, S.S. Chang, G. Dalbagni, et al. Guideline for the management of nonmuscle invasive bladder cancer (stages Ta, T1, and Tis): 2007 update. J Urol. 2007;178:2314-2330 Crossref.
-  M. Babjuk, W. Oosterlinck, R. Sylvester, E. Kaasinen, A. Böhle, J. Palou-Redorta. EAU guidelines on non-muscle-invasive urothelial carcinoma of the bladder. Eur Urol. 2008;54:303-314 Abstract, Full-text, PDF, Crossref.
-  J.W.F. Catto, M.F. Abbod, P.J. Wild, et al. The application of artificial intelligence to microarray data: identification of a novel gene signature to identify bladder cancer progression. Eur Urol. 2010;57:398-406 Abstract, Full-text, PDF, Crossref.
-  P.J. Wild, A. Herr, C. Wissmann, et al. Gene expression profiling of progressive papillary noninvasive carcinomas of the urinary bladder. Clin Cancer Res. 2005;11:4415-4429 Crossref.
Department of Urologic Surgery, Vanderbilt University Medical Center, A1329 Medical Center North, 1161 21st Ave. South, Nashville, TN 37232-2765, USA
© 2009 European Association of Urology, Published by Elsevier B.V.
Recommend this article
Currently this article has a rating of 0. Please log in to recommend it.