Protein folding atomistic models

The current understanding of the protein folding process has benefited much from studies that focus on computer simulations of simplified lattice models. These studies try to construct as simple a model as possible that will capture some of the more important properties of the real polypeptide chain. Once such a model is defined it can be explored and studied at a level of detail that is hard to achieve with more realistic (and thus more complex) atomistic models. 

In another study of protein structure, Cox and Johnston3 analyzed how the choice of GA parameters affects the quality of the GA search. This sort of approach was also adopted by Djurdjevic and Biggs,4 who presented a detailed study of how evolutionary algorithms can be used, in combination with a full atomistic protein ab initio model, for fold prediction and, like Cox and Johnston, considered the influence of the different values of parameters on the success of their protein folding calculations. 

There are immense challenges also to model protein function, which will rely on better theoretical models for secondary structure formation (a helices, p sheets). Models presently used are molecular force field approaches, which are rather phenomenological. Realistic atomistic modelling is a long-term goal. In the meantime, energy landscape approaches should help us elucidate the detailed folding mechanisms that lead to protein function. 

Fully atomistic simulations are the most realistic of the three simulation methods. They include a fully detailed description of the amino acids comprising the protein, and they are thus much more true to life than the other models. In addition, solvent molecules may be added explicitly or implicitly to the simulation. Because of this extreme detail, a simulation of a small protein may require the treatment of thousands of atoms. Fully atomic simulations are thus extremely computationally expensive, and only short time scales can be explored. As computational power continues to increase, so do the time scales accessible with this method. Nevertheless, fully atomic simulations still cannot capture kinetic information they are, however, useful in understanding important local interactions that drive protein folding. 

Usually, the mesoscopic, kinetic models are considered to be well suited for predicting dynamic properties of polymer solutions on macroscopic scales. Details of the fast solvent dynamics are in most cases irrelevant for macroscopic properties. Exceptions are polyelectrolytes, where the motion of counterions in the solvent can have a major influence on polymer conformation. Therefore, more microscopic models of polyelectrolytes with explicit counterions are sometimes employed [34] (see also the contribution by M. Muthukumar in this volume). Another exception is the dynamics of individual biopolymers, for example, protein folding, which is modeled with an all atomistic model including an explicit treatment of the (water) solvent molecules [35]. 

CG terms are calibrated on Potential of Mean Forces. These quantities can be extracted either from statistical distributions from the PDB data bank, or from atomistic MD simulations of model systems. Currently, the UNRES force field provides two sets of parameters. The first is calibrated over both structural and folding thermodynamics of G-related albumin-binding module (PDB code IGAB), a triple a-helix bundle. The second set of parameters is calibrated over folding data for tryptophan cage and tryptophan zipper 2 (PDB codes 1L2Y, ILEl), small protein constructs displaying both cx-helical and extended structural elements. 

---