Abstract
Recently, some studies have integrated the results of copy number alterations and analysis of transcripts aiming to identify mutations in genes that may be drivers on the tumor process. Similarly, several studies in cancer have been described using an integrated analysis between transcripts expression and methylation data. In this study it was developed a method that integrates genomic and transcriptomic alterations by detecting key mutations that drive cancer progression. The score-guided algorithm search for genes that are both recurrently aberrant in copy number and methylation data and associated with variance of expression patterns in tumor samples. This proposed method uses concepts derived from gene network inference and addresses a recurring problem in this approach. Usually, the main limitation of network inference methods is related to a small number of samples with large dimensionalities and the noisy nature of the expression measurements. To address these limitations, alternatives are necessary to obtain better accuracy on the inference problem. One of them, presented in this study, is the integration of different data in order to reduce errors during the gene prediction by feature selection methods based on conditional entropy for identification of gene network architectures. To search the key genes predictors, the score-guided algorithm is based in three databases aiming to select the best combination with the highest score, weighted by the gene expression matrix. This method may allow greater accuracy decreasing the error during the network prediction and selecting only relevant “hub” genes in the analysis. To test this score-based algorithm, it was used a 35 penile carcinomas as study model evaluated by large-scale transcript analysis, CGH-array and large-scale methylation data. Penile cancer is a rare neoplasm usually associated with low socioeconomic status, poor personal hygiene and HPV infection (particularly 16 and 18 subtypes in 40-50% of cases). The analysis correctly identifies known recurrent regions with copy number and methylation changes related to immune response and inflammation pathways. In addition, the “hub” genes in these regions were also connected to their associated aberrant neighbors. Since the penile carcinoma development is often attributed to inflammatory and irritative effects of smegma - a bioproduct of bacterial action in penile epithelial cells described as associated with this tumor - the present findings may contribute to a better understanding of the tumor development process by identification of key factors in the regulation of gene expression and genome stability. Thus the findings presented in this study demonstrate the ability of the integrative data algorithm to identify drivers as biomarkers in penile carcinomas.
Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr 3984. doi:1538-7445.AM2012-3984