Predicting antigenic determinants of foreign proteins from their amino acid sequence and/or conformation is of growing importance in the production of synthetic vaccines and antigens. Unlike foreign antigenic proteins, little is known of the suitability of predictive techniques for defining antigenic regions of self proteins recognised by autoantibodies. In this study we describe our use of two computer programmes (HYDRO 3 and ACROPHILICITY [ACRO], Hopp, 1986) for the prediction of antigenic determinants of autoantigens of the cell nucleus. Using the amino acid sequence of the protein, HYDRO 3 and ACRO respectively, provide information on the hydrophilic and surface regions of the protein. Both methods were used to predict the antigenic determinants of known autoantigens, including histones, the A, B", E and 70 kD proteins of snRNPs, SS-B/La, proliferating cell nuclear antigen (PCNA) and others. Our analysis of the antigenic determinants of histones agreed with other studies which have used antihistone antibodies and fragments of histones to show that autoantibody reactive sites reside in the terminal portions of these proteins, particularly the amino terminus. A detailed study of histone 2B correctly identified most regions recognised by antibodies, particularly autoantibodies. In addition the recently described epitope of the autoantigen ribosomal protein P2 was predicted by this analysis. From these observations we hypothesize that linear antigenic sites of self proteins can be predicted. Our hypothesis can be proven experimentally by demonstrating specific interaction between autoantibodies and synthetic peptides homologous with the predicted determinant.