BioProject

We analyzed the protein-coding and non-coding gene expression profiles of 64 samples of prostate cancer primary tumors. More...

We analyzed the protein-coding and non-coding gene expression profiles of 64 samples of prostate cancer primary tumors. All samples were collected between 1998 and 2001 with informed consent from patients subjected to radical prostatectomy at Hospital Sirio-Libanes in São Paulo. Selected patients were identified with clinical Stage T1-2 prostate cancer and no lymph node involvement, and received no adjuvant treatment after surgery as long as they remained recurrence-free. Biochemical recurrence was defined as an increase in patient blood PSA level to 0.2 ng per mL of blood at any time during the 5-year follow-up after prostatectomy. For this kind of experiment, also called self-self hybridization, the microarrays were cohybridized with each of Cy3- and Cy5-labeled cRNA replicates. This strategy has been used to derive intensity-dependent cutoffs to classify a gene as differentially expressed or divergent in comparative genomic hybridization (CGH) studies. The comparative analysis of constant fold change cutoffs and intensity-dependent ones has been extensively discussed, showing a superior performance of the intensity-dependent strategy. For the validation dataset processing, reference values obtained with the training dataset processin were applied to normalize the validation dataset. These values were: Average ranked intensities (quantile normalization), batch information (batch adjustment), and gene average and standard deviations (z-score transformation. Overall design: Here we describe the validation of the gene expression profile comprised of 32 protein-coding mRNAs and 6 intronic non-coding RNAs (ncRNAs) in an independent set of 22 samples, 16 from recurrence-free patients and 6 recurrent patients. In order to compare the expression levels of training and independent validation samples, gene intensity levels of samples in the validation dataset were transformed using normalization factors that had been generated with the training dataset. Less...