computer modeling in molecular biology.pdf

28 Tim Jl? Hubbard and Arthur M. Lesk2.3.2.3 Side Chain Building and Optimisation of Side ChainConformationOnce a complete main chain model has been constructed side chains need to be builtand their conformations determined.For closely-related proteins, it is observed that most side chains tend to retain conformation- even mutated ones. This is because each side chain, even those on thesurface, is packed in a cage formed by its neighbours. In closely-related proteins, amutated side chain is likely to find itself in a cage created largely by nonmutatedneighbours, and must conform itself to it. Therefore the first approximation shouldbe: for side chains that have not been changed from the parent structure, retain thesame conformation; for mutated side chains, retain the same conformation as faras the stereochemical similarity will allow. Of course application of this rule will producesome sterically impossible combinations.An essential step therefore is to adjust the side chain conformations to achievea low-energy conformation. There are now a large number of programs available tocarry out such building and packing automatically [37, 64, 65, 671 which are probablymore accurate (and of course much quicker) than manual manipulation [68].Currently available procedures for automatic side chain modelling and packingperform quite well when starting from experimental main chain atoms or Ca’s. Itis not so clear what happens as the position of the main chain in the model becomesless and less accurate (as the closeness of the relationship between the input sequenceand the sequence of known structure decreases). Exact Ca/main chain positions mayforce a unique side chain packing. In contrast, in a real modelling situation Cdmainchain positions will be inexact, and packing errors are more likely. In particular, theincorrect positioning of a large buried hydrophobic residue can result in seriouspacking errors within a whole region of the hydrophobic core of a protein.Finally, once all atoms have been built, the model can be subjected to EnergyMinimisation (EM) or Molecular Dynamics (MD). EM is a purely cosmetic operation.It will remove some bad atom contacts in a model but will not significantly altereven side chain conformations. It can be useful however to “clean up” a model bysmall local adjustments. MD can be used to explore a much greater conformationalspace around the model structure than EM. If the changes in conformation are tobe at all realistic, it is necessary to simulate in the presence of water. However, it isimportant to realise that if the starting model has substantial errors, even a very longMD run is very unlikely to improve it. Perhaps the most useful result of such simulationsis to observe how a model moves with time. If the model is wrong it is likelyto be more unstable when subjected to simulation.