Abstract
Much effort is invested in defining gene expression changes that can be monitored in surgical and biopsy samples of oral lesions to identify oral squamous cell carcinoma (OSCC), a cancer that strikes over 30000 people in the United States each year. There is an extensive list of genes shown to be differentially expressed in surgically obtained oral tumor tissue versus normal, non-malignant, tissue. These have been catalogued over the years and have the potential to inform OSCC characterization. At the same time noninvasive detection and diagnosis of OSCC is a goal that is being explored using a number of methods including brush oral cytology. Using RNA from oral cells obtained with brush cytology we tested levels of mRNAs of 22 genes shown to be differentially expressed in surgically obtained OSCC and normal tissue. Of these 10 showed differential expression in our training set of 12 tumor versus 18 non-malignant lesion samples all from tobacco users. Using Support Vector Machines an algorithm was developed that provided class prediction based on a subset 5 of these mRNAs. With the exception of one case, cross validation during training revealed that this OSCC classifier showed greater than 92% accuracy in differentiating OSCC from non-malignant oral lesions. We will report on its accuracy for correctly identifying OSCC samples in a validation set of RNA from brush cytology samples from an independent set of samples. This test of this methodology using RNA from brush cytology in identifying OSCC highlights the possibility of developing an RNA based classifier from brush cytology to screen for OSCC and perhaps pre-malignancies.
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 2099. doi:1538-7445.AM2012-2099
- ©2012 American Association for Cancer Research
Volume 72, Issue 8 Supplement
