Colorectal cancer (CRC) is a deadly disease that is often not diagnosed until the latter stages of development. For CRC patients therapeutic options include surgery and/or chemotherapy, choices that are most often driven by TNM staging methods that are not sensitive enough to accurately predict outcome. Therefore, more accurate predictors are needed to assist oncologists and surgeons in their treatment of patients with CRC, and molecular diagnostics offer the promise of providing such predictors. To identify molecular predictors of outcome in CRC patients we performed expression profiling experiments using Affymetrix oligonucleotide arrays in a training set comprised of primary tumor biopsy samples from 50 CRC patients. 24 patients in the training set were cancer free 3 years post-surgery and were designated as the “good prognosis” group. The remaining 26 patients have either distant metastasis at the time of surgery, or recurrence or cancer-related death within 3 year post-surgery, and were designated the “bad prognosis” group. Class prediction analysis between the two groups was performed and a set of marker genes was identified and cross-validated as a “prognosis signature”. The prediction accuracy of this set of genes was tested in a small, independent set of colon cancer tumors. The identified marker genes predicted CRC prognosis better than traditional methods and could direct treatment of CRC patients. Implications and functional annotations of prognostic marker genes in colon cancer progress and metastasis will be discussed.

[Proc Amer Assoc Cancer Res, Volume 45, 2004]