Individual bacteria and shifts in the composition of the microbiome have been associated with human diseases including cancer. To investigate changes in the microbiome associated with oral cancers, we profiled cancers and anatomically matched contralateral normal tissue from the same patient by sequencing 16S rDNA hypervariable region amplicons. In cancer samples from both a discovery and a subsequent confirmation cohort, abundance of Firmicutes (especially Streptococcus) and Actinobacteria (especially Rothia) was significantly decreased relative to contralateral normal samples from the same patient. Significant decreases in abundance of these phyla were observed for pre-cancers, but not when comparing samples from contralateral sites (tongue and floor of mouth) from healthy individuals. Weighted UniFrac principal coordinates analysis based on 12 taxa separated most cancers from other samples with greatest separation of node positive cases. These studies begin to develop a framework for exploiting the oral microbiome for monitoring oral cancer development, progression and recurrence.

We demonstrate significant predictive performance of the risk model in a new cohort of patients with primary HNSCC, adjusted for confounders. Our training experience also supports the feasibility of adapting the risk model in clinical practice.

We applied a robust combinatorial (multi-test) approach to microarray data to identify genes consistently up- or down-regulated in head and neck squamous cell carcinoma (HNSCC). RNA was extracted from 22 paired samples of HNSCC and normal tissue from the same donors and hybridized to the Affymetrix U95A chip. Forty-two differentially expressed probe sets (representing 38 genes and one expressed sequence tag) satisfied all statistical tests of significance and were selected for further validation. Selected probe sets were validated by hierarchical clustering, multiple probe set concordance, and target-subunit agreement. In addition, real-time PCR analysis of 8 representative (randomly selected from 38) genes performed on both microarray-tested and independently obtained samples correlated well with the microarray data. The genes identified and validated by this method were in comparatively good agreement with other rigorous HNSCC microarray studies. From this study, we conclude that combinatorial analysis of microarray data is a promising technique for identifying differentially expressed genes with few false positives.

No abstract

No abstract