Homology modelling by distance geometry

Abstract

Unknown protein structures can be predicted from known structures (the scaffolds) with sequences sufficiently homologous to that of the target, based on the observation that similar sequences usually adopt the same fold. When structural equivalences between residues in the scaffold and target proteins are expressed in terms of conserved interatomic distances, the resulting 'distance geometry' representation provides an elegant mechanism for simultaneous restraint satisfaction and bias-free conformation space exploration. We present a homology modelling algorithm based on distance geometry that relies on the gradual projection of simple model chain coordinates into Euclidean spaces with decreasing dimensionality. The similarity between the unknown target structure and the scaffold proteins with known structures was described by mapping secondary structure assignments and specific distance restraints between C alpha atoms onto the model through a multiple alignment. This information was complemented by additional restraints derived from stereochemical considerations and other general aspects of protein structure such as hydrophobic core formation or the absence of tangled mainchains. The method was capable of quickly locating the correct fold even from an alignment with modest average conservation indicating that it could serve as a fast tool for obtaining correct low-resolution starting conformations for detailed refinement.