Purpose: To select autoantibody signatures for early detection of colorectal cancer (CRC). Experimental Design: A phage cDNA expression library was constructed with fresh tumors from 30 CRC patients and biopanned by using serum pools of 20 CRC patients and 20 healthy controls. A classifier was discovered in the training set of 30 CRC patients at stages I and II and 30 matched healthy controls and then blindly validated in an independent set of 60 CRC patients, 60 healthy controls, 52 polyps patients, and 30 autoimmune diseases patients. Expression of proteins was examined by using immunohistochemistry.Results: Five-phage peptide clones showing higher discriminatory power than others in training set were selected for validation. The five-phage peptide classifier was able to discriminate between early CRC patients and healthy controls, with sensitivities of 90.0% to 92.7% and specificities of 91.7% to 93.3%. In those with serum carcinoembryonic antigen less than 5 ng/mL, the classifier was efficient in discriminating CRC from healthy controls, with an area under the curve of 0.975. The classifier was able to discriminate all of the 9 patients with serrated adenoma from healthy controls. Thirteen (43.3%) of the patients with autoimmune diseases were misclassified. Of the five phage peptides, one encoded a peptide identical to immunoglobulin G (IgG) heavy-chain constant region. IgG immunostaining was stronger in mesenchymal cells than in cancer cells in the tumors and was apparent in serrated adenoma.Conclusions: The five-phage peptide classifier stands out as promising early diagnostic biomarkers for CRC, but it is unsuitable for discriminating CRC from autoimmune diseases. Truncated IgGs generated from the tumors might be novel CRC-associated antigens. Clin Cancer Res; 17(17); 5715-24. Ó2011 AACR.