Abstract
Abstract: :
Purpose: We have shown by gene expression profiling that uveal melanomas cluster into two prognostic groups. Class 1, the less aggressive class is made up of spindle cells whereas the second more aggressive class is morphologically epithelioid. This study was performed to determine if the prognostic molecular classification could be extended to a clinically feasible assay that could be performed on a fine needle biopsy specimen. Methods: We isolated RNA from uveal melanoma tumor samples and hybridized it to Affymetrix gene chips, HU 133A and B. Using weighted voting and support vector machine algorithms, all samples were assigned to the correct tumor class. In order to confirm the results we saw in the microarray data we used ∼40mg of tumor from each of 13 known and 8 unknown samples and analyzed them using qRT–PCR. Results: We first compared gene expression profiles to tumor cytology as determined by independent pathological reports, and saw that they naturally fell into two classes. The results from the PCR assay segregated the tumors into the same two classes. We then compared the profiles of the unknown samples to the known samples. As expected, the eight unknown samples divided into the proper classes when compared to independent pathological reports. Conclusions: We can confidently predict which class a tumor will fall into using microarray analysis with as few as four genes. This assay can be used with small amounts of RNA as from needle biopsies to predict the patients’ tumor class and survival. Potentially, this information can be used to tailor treatment regiments to the needs of each individual case.
Keywords: melanoma • gene microarray • gene screening