Self tolerance to MHC class I-restricted nonmutated self Ags is a significant hurdle to effective cancer immunotherapy. Compelling evidence is emerging that altered peptide ligands can be far more immunogenic than their corresponding native epitopes; however, there is no way to reliably predict which modifications will lead to enhanced native epitope-specific immune responses. We reasoned that this limitation could be overcome by devising an empirical screen in which the nearly complete combinatorial spectrum of peptides of optimal length can be rapidly assayed for reactivity with a MHC class I-restricted cytotoxic T cell clone. This method, solid-phase epitope recovery, quantitatively ranks all reactive peptides in the library and allows selection of altered peptide ligands having desirable immunogenic properties of interest. In contrast to rationally designed MHC anchor-modified peptides, peptides identified by the present method are highly substituted in predicted TCR contact residues and can reliably activate and expand effector cell populations in vitro which lyse target cells presenting the wild-type epitope. We demonstrate that solid-phase epitope recovery peptides corresponding to a poorly immunogenic epitope of the melanoma Ag, gp100, can reliably induce wild-type peptide-specific CTL using normal donor T cells in vitro. Furthermore, these peptides can complement one another to induce these responses in an overwhelming majority of normal individuals in vitro. These data provide a rationale for the design of superior vaccines comprising a mixture of structurally diverse yet functionally convergent peptides.