Evaluation of microarrays for measuring cell cycle progression (CCP) gene expression.

Abstract

e16566 Background: A CCP score has been developed and validated to provide prognostic information in prostate cancer. To date, our studies have exclusively used qRT-PCR to measure CCP gene expression, which is generally considered the ‘gold-standard’ for measuring RNA expression levels. However, qRT-PCR allows for the interrogation of a relatively small number of genes. Because prostate cancer biopsies provide very limited tissue, there is interest in evaluating RNA expression platforms that could simultaneously integrate CCP genes and other less characterized targets that might be clinically useful. Here we evaluate the ability of microarrays to measure CCP gene expression. Methods: CCP scores for 1636 radical prostatectomy samples submitted for commercial testing were generated from the mean expression of 31 CCP genes and 15 housekeeping genes as determined by qRT-PCR. CCP scores were also generated from several publically available Affymetrix microarray expression data sets (Nakagawa et al. 2008, Karnes et al. 2013, Klein et al. 2014, Boormans et al. 2013, Taylor et al. 2013) by averaging the intensities (RMA processed, base 2 log scale) of the 31 CCP genes. Expression data were evaluated by comparing CCP score range and pairwise correlations between CCP genes. Results: The average pair-wise correlation in the commercial cohort using qRT-PCR was 0.67. The pair-wise correlations in the microarray studies were significantly lower, ranging from 0.17 (Klein) to 0.58 (Boormans). Importantly, the data in Boormans et al. was generated from frozen tissue. The CCP score in our commercial cohort ranged from -2 to 3 (base 2 log scale) with a standard deviation (SD) of 0.79. In contrast, the CCP score range in microarray studies was highly truncated with SDs ranging from 0.12 (Nakagawa) to 0.46 (Boormans). Conclusions: Expression of CCP genes as determined by microarrays correlates weakly with expression measured by qRT-PCR, with reduction in normally observed pairwise correlations among CCP genes and the range of CCP scores. This is especially true if the RNA is derived from fixed tissue. As a result, microarray generated CCP scores cannot be assumed to be a valid surrogate for qRT-PCR generated scores for prediction of patient outcome.