Side chain conformations search

C Wilson, LM Gregoret, DA Agard. Modeling side-chain conformation for homologous proteins using an energy-based rotamer search. J Mol Biol 229 996-1006, 1993. 

Considering only the conformations with a total energy less than 10 kJ mol" above the lowest energy conformations, the same ensemble of conformations as detected in the structural database searches was found. With this systematic backbone and side chain rotamer search approach, we analysed only 144 structures instead of the 10 structures for a conventional grid search with 36° intervals for 6 dihedral angles. 

Side-chain conformation prediction is a combinatorial problem, since there are on the order of n ot possible conformations, where nrot is the average number of rotamers per side chain and N is the number of side chains. But in fact the space of conformations is much smaller than that, since side chains can only interact with a small number of neighbors, and in most cases clusters of interacting side chains can be isolated and each cluster can be solved separately [93, 165]. Also, many rotamers have prohibitively large interactions with the backbone and are at the outset unlikely to be part of the final predicted conformation. These can be eliminated from the search early on. 

Many standard search methods have been used in side-chain conformation prediction, including Monte Carlo simulation [176-178], simulated annealing [179], self-consistent mean field calculations [154, 173, 180], and neural networks [170]. Self-consistent mean field calculations represent each side chain as a set of conformations, each with its own probability. Each rotamer of each side chain has a certain probability, p(n). The total energy is a weighted sum of the interactions with the backbone and interactions of side chains with each other ... 

The latest developments in modeling are position-specific rotamer searches (Fig. 2). Dunbrack et al. and Vriend et al. showed that the preferred rotamer is to a large extent determined by a local backbone [34,35], Database searches for peptides with the same local backbone and the same amino acid type in the middle often reveal a more restricted pattern of side-chain conformations than one would expect from generalized rotamer libraries. Position-specific rotamer distributions are obtained by searching the database for segments of residues that fulfill two criteria ... 

Another way to simplify the tertiary structure problem is to fix the backbone and then carry out an exhaustive search on the allowed side chain conformations. Desmet and co-workers have developed a dead-end elimination method for searching side chain conformations. Side chain conformations are grouped into a limited set of allowed rotamers. While an exhaustive search of all possible combinations of these rotamers is still not feasible, the application of the dead-end elimination theorem allows removal of impossible combinations early in the search, thus controlling the combinatorial explosion and leading to a small group of possible final solutions. The possible solutions can then be compared to find the best possible structure. 

Critical Comparison of Search Algorithms Applied to the Optimization of Protein Side Chain Conformations. 

The O-C-C-O-H side-chain conformation is CiGG (gauche,gauche) in all cases. The most stable hydroxyl conformation was determined for each skeletal conformer by conformational search (Appendix) and was used for these calculations. 

A different branch-and-bound approach, based on an inequality called the dead-end elimination theoremhas been developed in the context of determining side chain conformations in proteins, and will be briefly discussed with other side chain search methods. 

Side chain generation is often a source of error. It will be most reliable if certain rules of thumb are obeyed. Start with structurally conserved side chains and hold them fixed. Then look at the energy and entropy of rotamers for the remaining side chains. Conventional conformation search techniques are often used to place each side chain. 

Use conventional conformation search techniques to optimize side chains. 

Perform a conformation search of the protein backbone using the meso-scale side-chain representation. 

We have added a companion option to PBUILD, PRANDOM which eases considerably the problem of finding good conformations of a polymer segment. PRANDOM automatically selects all of the polymer backbone and/or side chain bonds and will randomly select rotations for each bond. In a few minutes, one can not only build a polymer fragment, but also set up a Monte-Carlo search of its conformational space. However, even this cannot solve the problems for large models (pentamer or larger), again due to the number of bonds to be rotated. 

---