Profiling tumor-associated autoantibodies for the detection of colon cancer.

PURPOSE The purpose of the present study was to screen the autoantibody signature of colon cancers to develop serum markers for colon cancer detection. EXPERIMENTAL DESIGN A phage cDNA expression library of colon cancer was built. The library was sequentially screened by a pool of 10 colon cancer sera, goat antihuman IgG, and a pool of two healthy sera to identify phage-expressed antigens recognized by tumor-associated antibodies. The clones picked out by these screening were subjected to a training set with 24 colon cancer sera and 24 healthy sera. The antigen combination, which got the most satisfactory classification, was tested by an independent set of 24 colon cancer sera with equal number of sera from normal donors. The carcinoembryonic antigen (CEA) level of these sera was detected for the additional classification analysis with or without the antigen combination. RESULTS A cDNA expression library consisting of 2 x 10(6) primary clones was prepared. After three turns of screening, 24 antigens recognized by tumor-associated antibodies were picked out for serum marker identification. The training set showed that a six-marker combination got the most satisfactory classification in a logistic regression model; leave-one-out validation achieved 91.7% sensitivity and 91.7% specificity. In a testing set with this marker panel, we correctly predicted 85% of the samples. Although according to CEA level alone, we correctly predicted 75% of the samples with 42% of cancer patients misclassified. When CEA was combined with the six markers, the sensitivity and specificity increased to 91.7% and 95.8%, respectively. The six antigen sequences in the phage display system are relatively short peptides. Only two of them showed homology to known protein sequences. CONCLUSIONS Autoantibodies against phage-expressed antigens derived from colon cancer tissues could be used as serum markers for the detection of colon cancer.