Creating structure features by data mining the PDB to use as molecular-replacement models

Abstract

Mathematical data-mining techniques to generate a representative set of protein fragments are described. Protein fragments are used as search models within the macromolecular phasing method of molecular replacement to attempt to phase protein data without a homologous model correctly. Preliminary investigations using these fragments indicate that molecular replacement with AMoRe is not sensitive enough to phase myoglobin or insulin data sufficiently for successful refinement. The results suggest that more advanced molecular replacement techniques may be successful, though at present these are not computationally practical.