Designing rotamer libraries based on pairs of consecutive residues: A preliminary analysis

Abstract

Predicting with high accuracy the packing of side chains in proteins has been recognized as a tough challenge. Approaches that tackle this problem have three basic elements: the library of rotamers, the energy function, and the search algorithm. Libraries of rotamers have been developed based on statistical analysis of side chain's conformations in known, high quality protein structures. Rotamer libraries are becoming better with respect to torsion angles coverage, at the expense of becoming larger and more complex. Larger libraries impact on the algorithm's search space size. The size of the algorithm's search space is bounded by O(bn) with n the protein length, and b the maximum number of rotamers among all amino acids. It is clear that as n increases the search space size will exponentially increase. Motivated by this observation, the feasibility of designing rotamer libraries which consider pairs of consecutive residues at the same time, and that consequently reduces the exponent of the search space's size from n to n/2, is analyzed here. The proposed analysis works as follows, given a set of target protein structures, and a value for a desired coverage level of dihedral angles, for each pair of contiguous residues in the target set, computes the set of rotamers needed to achieve the specified coverage level. As a proof of concept the proposed method uses a set of 149 protein structures to generate the library. The experimental results show that: i) the number of required rotamers of pairs of residues grows slowly with the number of required rotamers in a standard library for the same coverage level, and ii) the achievable accuracy level of the generated library is high.