BioProject

Samples were prospectively collected from patients with histologically normal surgical resection margins. More...

Samples were prospectively collected from patients with histologically normal surgical resection margins. 96 tissue samples (histologically normal margins, oral carcinoma and adjacent normal tissues) from 24 patients comprised the training set. Our study design was guided by the hypothesis that the expression of genes present in oral squamous cell carcinoma (OSCC) but not in healthy oral tissues would be indicative of recurrence in advance of histological alteration. We used meta-analysis of five published microarray data sets (GEO accession GDS2520, Kuriakose et al. 2004; GDS1584, Toruner et al. 2004; GSE6791, Pyeon et al. 2007; GSE9844, Ye et al. 2008; and GSE10121, Sticht et al. 2008), in conjunction with the current training set, to identify genes reliably over-expressed in OSCC. This reduced gene set was used to train a risk model to predict recurrence based on over-expression of a subset of these genes in histologically normal surgical resection margins. Validation of the risk signature was performed using quantitative real-time reverse-transcription PCR in an independent set of 136 samples from an independent cohort of 30 patients. Overall design: This was a case-only design involving a training set of 23 tumors and 73 margins from 24 patients with squamous cell carcinoma of the tongue. Less...