HIV-1 genetic diversity is a major obstacle for developing an effective vaccine. My hypothesis is that HIV-1 genetic diversity can be characterized and that cross-clade immunogens can be predicted at the population level. I systematically investigated positive selection (PS) pressures on HIV-1 Env and Gag proteins based on the analysis of the sequences collected from the Los Alamos Sequence Database. I identified PS sites, investigated PS patterns, correlated PS with the known functional sites of the two proteins, calculated frequencies of HLA alleles targeting CTL epitopes, and compared PS patterns among major subtypes. The results showed that PS pressure was widely dispersed across the entire regions of both HIV-1 Env and Gag proteins, suggesting the conserved regions are under host immune response pressure. The neutralizing antibody, non-neutralizing antibody, and CTL responses were found to be the major forces driving genetic diversity of HIV-1 env and gag genes at population level. However, PS pressures on both Env and Gag proteins remain stable over time, suggesting genetic diversity of HIV-1 driven by host immune responses changed very little over the last 29 years. Furthermore, the results also demonstrated that up to 70% PS sites were shared among the major HIV-1 clades, implying the existence of cross-clade immunogenicity. A number of potential cross-clades immunogens were predicted to elicit CTL or neutralizing antibody responses from Env and Gag proteins. I also detected a significant correlation between HLA allele frequencies and host CTL responses elicited by Accessory/Regulator’s proteins at population level. Moreover, I detected an association between the frequency of HLA-B7 supertype and the number of identified optimal CTL epitopes. The results suggest HLA class I allele frequencies in a population influence the evolution of HIV-1. I also systematically evaluated the utility of ultra-deep pyrosequencing to characterize genetic diversity of HIV-1 gag genes within quasispecies. The results showed that ultra-deep pyrosequencing of amplified HIV genes is a better method than the traditional Sanger-clone-based method in the comprehensive characterization of genetic diversity of HIV-1 quasispecies, especially in detecting low frequency variations. In conclusion, my thesis provides important information for rational design of an effective HIV-1 vaccine.