The full protein coding region of human immunodeficiency virus (HIV) genomes were sequenced using plasma collected from nine African-Americans prior to seroconversion and 7 to 28 days later. HIV mutations emerged in seven of these subjects at a genomewide rate of 2% per year. The location of nonsynonymous (NS) HIV mutations within these subjects was compared to their potential HLA-A and B types restricted CTL epitopes reported in the Los Alamos National Laboratory HIV immunology database. A statistically significant (P < 0.005) number of the early NS mutations (13.5%) were found within previously reported CTL epitopes. A virus sequencing and reported CTL epitopes database analysis therefore support a model where a significant proportion of very early nonsynonymous HIV mutations are selected by CTL.