Genetic Algorithms and Protein Folding

1.2.1.6 Side Chain Placement

Crystallographers often face the problem of positioning the side chains of a protein when the primary structure and the conformation of the backbone is known. At present, there is no method that automatically does side chain placement with sufficiently high accuracy for routine practical use. Although the side chain placement problem is conceptually easier than ab initio tertiary structure prediction it is still too complex for analytical treatment.

The genetic algorithm approach as described above can be used for side chain placement. The torsion angles , , and have simply to be kept constant for a given backbone. Side chain placement by the genetic algorithm was done for Crambin. For each five residues, a superposition of the native and predicted conformation is shown in stereo projection graphs in Figure 5. As can be seen, the predictions are quite well in agreement with the native conformation in most cases. The overall r.m.s. difference in this example is 1.86Å. This is not as good as but comparable to the results from a simulated annealing approach [31] (1.65Å) and a heuristic approach (1.48 ) [32].

Figure 5. Side Chain Placement Results

TTCCP SIVAR SNFNV CRLPG TPEAI CATYT GCIII PGATC PGDYA N

It must be emphasised that these runs were done without optimising either the force field parameters of the fitness function or the run time parameters of the genetic algorithm. From a more elaborate and fine-tuned experiment even better results should be expected.