Emily Cannon1, Steven Buechler2. 1. Department of Applied and Computational Mathematics and Statistics, University of Notre Dame, Notre Dame, IN. 2. Department of Applied and Computational Mathematics and Statistics, University of Notre Dame, Notre Dame, IN. Electronic address: steve@nd.edu.
Abstract
BACKGROUND: Cancers of the right colon have been shown to differ from left-side colon cancers in prognosis, response to epithelial growth factor receptor inhibitors, microsatellite instability and BRAF mutation status, and other molecular characteristics. Clinical application of these differences will benefit from a deeper understanding of how tumor location defines and is defined by gene expression. MATERIALS AND METHODS: This study was carried out using Affymetrix microarray datasets (Cohort A: training set, n = 352; validation set, n = 519) and samples from The Cancer Genome Atlas Colon Adenocarcinoma database (Cohort B: n = 408), in which tumor location was reported. Gene expression patterns characteristic of tumor side were identified in a manner unbiased by statistical classification method. RESULTS: In the Cohort A validation set, the anatomic locations of 75% of tumors agree with the locations predicted by gene expression (so-called genomic location), whereas 8% of tumors had genomic locations discordant with their anatomic locations, and 17% of tumors had ambiguous genomic locations. Genomic location was a better predictor of microsatellite instability, CpG island methylator phenotype status, and BRAF mutation status than anatomic location. Tumors with ambiguous genomic location were significantly (P = 1.3 × 10-7) more likely to have the mesenchymal consensus molecular subtype (40%) than those with a specific genomic location (18%). A genomic signature to predict genomic location was defined. CONCLUSION: Tumor location is increasingly considered in deciding treatment of a colon tumor. We showed that genomic location was superior to anatomic location as a predictor of molecular characteristics, suggesting that it may be a more accurate predictor of response.
BACKGROUND:Cancers of the right colon have been shown to differ from left-side colon cancers in prognosis, response to epithelial growth factor receptor inhibitors, microsatellite instability and BRAF mutation status, and other molecular characteristics. Clinical application of these differences will benefit from a deeper understanding of how tumor location defines and is defined by gene expression. MATERIALS AND METHODS: This study was carried out using Affymetrix microarray datasets (Cohort A: training set, n = 352; validation set, n = 519) and samples from The Cancer Genome Atlas Colon Adenocarcinoma database (Cohort B: n = 408), in which tumor location was reported. Gene expression patterns characteristic of tumor side were identified in a manner unbiased by statistical classification method. RESULTS: In the Cohort A validation set, the anatomic locations of 75% of tumors agree with the locations predicted by gene expression (so-called genomic location), whereas 8% of tumors had genomic locations discordant with their anatomic locations, and 17% of tumors had ambiguous genomic locations. Genomic location was a better predictor of microsatellite instability, CpG island methylator phenotype status, and BRAF mutation status than anatomic location. Tumors with ambiguous genomic location were significantly (P = 1.3 × 10-7) more likely to have the mesenchymal consensus molecular subtype (40%) than those with a specific genomic location (18%). A genomic signature to predict genomic location was defined. CONCLUSION:Tumor location is increasingly considered in deciding treatment of a colon tumor. We showed that genomic location was superior to anatomic location as a predictor of molecular characteristics, suggesting that it may be a more accurate predictor of response.