Colorectal cancer is the second most common cause of cancer death. Patients with less-advanced disease (stages l-lll) are either observed only or treated with adjuvant therapy to reduce the chance of recurrence after surgery. Currently, the decision to recommend adjuvant therapy is based almost entirely on conventional clinic pathological data. Most stage III patients are recommended to receive a fluorouracil based chemotherapy, whereas most stage l-ll patients are not. However, limitations in the prognostic power of the clinic pathological data alone likely leads to the unnecessary treatment of stage III patients who do not need it and the lost opportunity for treatment of stage II (and, potentially, stage I) patients that would benefit. In addition, a significant percentage of stage II patients opt to receive chemotherapy despite the marginal evidence of benefit. An accurate prognostic test would help guide patients with high-risk disease to choose chemotherapy and improve the quality of life of low-risk patients by sparing them from the side effects. This proposal describes the innovative usage of multiple machine learning methodologies as part of a comprehensive nonlinear model approach to predicting colorectal cancer recurrence based on a combination of standard clinic pathological data and data from multiple, carefully selected molecular markers. The ultimate product will consist of a panel of assays for the relevant molecular markers and the associated prognostic algorithm that predicts recurrence. Colorectal cancer patients and their oncologists will be able to use the results to guide their treatment decisions, resulting in positive impacts on life expectancy and quality of life.