A Cluster Analysis on the Structural Diversity of Protein Crystals, Exemplified by Human Immunodeficiency Virus Type 1 Protease

Information on many protein crystal structures has recently become available due to developments in crystallographic techniques. Even for a single kind of protein, several and sometimes many crystal structures are available. Human immunodeficiency virus type 1 (HIV-1) protease is one of the most extensively studied viral proteins, and about six hundred crystal structures have been determined. In this work, we examined the structural diversity of HIV-1 protease, classifying crystal structures into several groups from the viewpoint of similarity in atom geometry. Using 499 crystal structures downloaded from the Protein Data Bank (PDB), cluster analysis was applied to the whole body of HIV-1 protease and also to a limited number of residues at the binding pocket. As a consequence of clustering with regard to the whole body, 499 crystal structures were separated into 6 groups. It was found that a major factor for this separation is the space group of the crystals and that the space group strongly depends on the agents used in the protein crystallization. Amino acid mutation is a minor factor for separation in clustering. In cluster analysis for a limited number of residues at the binding pocket, crystal structures were not distinctly separated, and no clear factor linked to the separation was clarified. The results suggest that amino acid mutations have little effect on the coordinates of the main-chain atoms of HIV-1 protease. Hence, the changes in drug efficacy or substrate fitness caused by mutations are mainly due to the physicochemical features of amino acid side chains.