One of the greatest challenges in HIV vaccine development is accommodating the worldwide sequence diversity of the HIV-1 virus. To understand how viral sequence diversity may affect the potential breadth of HIV-1 vaccines designed to elicit antiviral T cell immunity, we have developed novel approaches to assess sequence conservation at the amino acid level, where vaccine effects are exerted. Taking each sequence from the LANL 2004 amino acid alignments as a potential vaccine or as a challenge virus, all pairwise combinations of sequences were evaluated by two methods: first, a traditional comparison of aligned sequences, and second, by a new walking 9-mer algorithm chosen to emphasize the typical length of an MHC-I epitope. The rules for comparing mismatched 9-mer pairs between vaccine and challenge sequences were empirically deduced from an experiment on Nef-specific CD8 epitopes and the viral sequences from naturally HIV-1-infected patients. Results were weighted such that each clade contributed in proportion to its global prevalence. Cross-clade breadth of response is best maintained for vaccines encoding Pol and Gag, while commonly proposed Env- and Tat-based vaccines would be more clade sensitive. We evaluated the additional breadth that could be expected from multiclade vaccines including consensus and ancestral sequences. For more diverse proteins, adding a second strain can add a significant increase in breadth, although for three or more strains the intrinsic diversity of the protein leads to diminishing improvement.