Selection from phage antibody libraries can be considered to be an in vitro immune system in which the antibody response is reduced to the bare minimum of antigen recognition. Using selections of antibodies on peptides from a phage antibody library, we investigated what constitutes peptide antigenicity in the context of the antibody-protein binding site. We selected polyclonal antibodies in a high throughput format against 44% of 90 overlapping peptides derived from three different proteins. Of these, 33% of peptides (epitopic peptides) were able to select antibodies that recognized the protein from which the peptides were derived. Although no algorithm was able to predict all epitopic peptides, solvent accessibility was the best predictor in this cell-free antibody selection context. We subsequently applied solvent accessibility to successfully predict epitopic peptides from p53 and Znf217, and showed that such peptide selected single-chain antibodies were able to recognize soluble p53 in ELISA and Znf217 in a western blot. This is likely to have considerable utility in functional genomics and proteomics where it should be possible to select antibodies against gene products on the basis of deduced amino acid sequence in a high throughput fashion
