We　have　developed　a　three-dimensional　pharmacophore　model　for　the　human　immunodeficiency　virus　type　I　(HIV-1)　integrase　(IN)　from　diketoacids　(DKAs)　inhibitors　using　the　genetic　algorithm　similarity　program　(GASP).　For　the　selected　training　set,　reliable　drug-like　properties　and　DKA-like　pharmacophore　features　exist.　Inhibitor　conformations　were　mapped　into　the　pharmacophore　model　and　superimposed　in　their　docking　conformations.　Corresponding　positions　between　the　pharmacophore　model　and　IN　residues　were　thus　obtained.　The　pharmacophore　model　was　refined　according　to　whether　the　pharmacophore　features　were　compatible　with　residues　around　them.　Finally,　an　optimal　pharmacophore　model　was　generated　and　consisted　of　I　hydrophobic　feature,　3　hydrogen　pair　features　and　I　hydrogen-bond　donor　feature.　The　pharmacophore　model　had　higher　reliability　with　a　goodness　of　hit　(GH)　score　of　0.56,　a　high　percentage　yield　of　actives　(1)　of　63.6%　and　a　lower　false　positive　rate　(FP)　of　0.41　%.　This　pharmacophore　model　can　contribute　to　the　discovery　and　design　of　new　DKA-like　inhibitors.