We have developed an algorithm to predict protein antigenic sites recognized by T lymphocytes, based on the finding that a majority of immunodominant T-cell sites tend to form amphipathic alpha-helices. We have conducted a thorough examination of hydrophobicity scales and computational procedures useful in detecting potential amphipathic helices. We also have developed a powerful statistical procedure based on Monte Carlo methods that evaluates physical-chemical properties proposed to be characteristic of T-cell antigenic sites. The computer program implementing the procedure compares any proposed property as expressed in the known T-cell sites with its expression in comparable randomly chosen sites. Other characteristics of known T-cell sites with potential predictive value are being sought. The methods developed are not unique to T-cell antigenic sites, but are useful in identifying physical- chemical characteristics of many biomolecular features (antibody binding sites, enzyme active sites, DNA promoter regions, for example). (See also report Z01 CB 04020-09 MET of Jay A. Berzofsky, who initiated many of the questions and with whom this work is closely coordinated.)
