Statistical modelling of globular proteins

Protein structure simulations are important for understanding and exploring properties of proteins and evaluating algorithms in bioinformatics. For example, computer-generated protein structures designed to mimic real a protein, decoys can be used to test the validity of a protein model. The model is considered correct only if is able to identify the native state configuration of the protein among the decoys. For example, the computer model being tested will be used to calculate the free energy of the protein in the decoy configurations. The minimum requirement for the model to be correct is that it identifies the native state as the minimum free energy state. A similar use of simulated protein structures or decoys is to compare or evaluate matching algorithms at different ground truth settings. Also in protein folding simulations, decoys are used to overcome the problem presented by the enormity of the conformational space. For very detailed protein models, it can be practically impossible to explore all the possible configurations to find the native state. To deal with this problem, one can make use of decoys. The idea behind this, is that the native configuration has not to be blindly searched through all possible conformations; the search can be limited to a relevant sub-set of structures. To start with, all non-compact configurations can be excluded