21042 Introduction: Pathologic stage is the most powerful predictor of outcome in colorectal cancer. However, in early stage disease there continues to be a need for better identification of patients at high risk of disease progression after surgical resection. We have undertaken a large scale effort translating gene expression profiles into immunohistochemical biomarkers that classify carcinoma in novel ways. Herein we report the identification of a colorectal tailored antibody “panel of diversity” that distinguishes tumor heterogeneity and propose a multivariate index assay (MIA) for predicting outcome. Methods: 744 stage I-III consecutive colon cancer patients treated by the CCIH between 1992 and 2000 were identified. Clinical followup through 2005 was obtained from the tumor registry and verified by chart review . Tissue micorarrays (TMA) were constructed from these patients resected primary tumors. Over 700 antibodies targeted by carcinoma gene expression experiments or literature review were first screened on a 30 case TMA. Of these, 34 were selected as showing subjectively high quality stains and the ability to classify patient populations. These antibodies were used to stain the clinical TMA and scored using a semi-quantitative scale. We used a number of approaches including Cox, RPART, and Bayesian to derive candidate MIA's to predict recurrence. Results: 13 antibodies showed a significant or near significant association with either overall recurrence or recurrence at 5 years. We have previously shown these antibodies to have similar prognostic abilities in other solid tumor carcinomas. We propose a model combining nine markers (p53, CCNA2, TFF3, RERG, AKR1C1, TLE3, IRX3, SYP, TTC7) as an MIA for predicting outcome in Stage II patients. The model is independent of pathologic stage. Conclusions: This colon tailored antibody ‘panel of diversity’ is a tool for characterizing the biologic diversity in colon cancer cohorts. The identification of nine prognostic markers that can be effectively combined using an MIA suggests that combination of multiple markers will be useful for developing sensitive and specific prognostic assays. The model proposed herein should be validated using additional cohorts to evaluate its overall clinical utility. No significant financial relationships to disclose.