Identification of native conformations of proteins using a low resolution model

Abstract

Exhaustive enumeration is performed by using a low resolution model. The secondary structure elements of the proteins are kept as rigid blocks. Large ensembles of decoy structures are generated by rotating the virtual flexible bonds. Several constraints are used to ensure the diversty and proper sampling of native-like conformations. These include the excluded volume requirement, the radius of gyration constraint, and the need for the occurrence of a sufficiently strong attractive interaction to hold the amino acids in a stable, coherent form. The number of accepted conformations is reduced to 1500-5000 decoys for each protein by using these contraints. To distinguish the most native-like fold, two criteria are used: Root-mean-square (RMS) deviations with respect to x-ray structure, and energies of the generated conformations. Energies are evaluated on the basis of knowledge-based potentials. The RMS deviation between the lowest energy conformation and those of x-ray structures is found to remain lower then 2.0 A. This observation indicates the suitability of the low resolution model and parameters for distinguishing the native fold. The dihedral angle preferences of the flexible bonds in the set of low energy conformations are examined by obtaining the torsional state probability distribution curves.