Physical Model
Rotamers and the Fixed Backbone
In order to represent the physical world in a computer, a number of approximations must be made. These approxmiations vary between applications. For instance, what might be important in a molecular dynamics or protein folding simulation may not be the same as what is important in a protein design calculation. Because of this, no single set of approximations is appropriate for all problems.
The most common approximations used in protein design are rotamers and the fixed backbone. Each of these are important to understand before describing how to set up a protein design application.
Rotamers
In a real protein, side-chains can sample a continuous space of conformations by rotating about their bonds. Rotable bonds in a side-chain are called chi-bonds and their corresponding angles of rotation are called chi-angles. Arginine, displayed below, has four indicated chi-bonds.
Computers are not very good at treating continuous spaces. Technically, there exists an infinite number of possible conformations a sidechain could adopt. Luckily, there are some restrictions on these conformations. Due to steric clashes, not all angles are energetically favorable. Also, when searching crystal structures in the Protein Data Bank, sidechains tend to adopt a limited range of conformations.
This allows us to define representative conformations, called rotameric conformations (or rotamers for short), that discretize the continuous space into a smaller set of unique structures. Arginine, for instance, could be represented by a collection of 100 rotamers, each representing a potential conformation it adopts in a protein. Alanine, on the other hand, would not need nearly as many rotamers to model its allowed conformations.
Fixed Backbone
Another problem protein design applications face is the possibility of backbone movement. Mutating the sequence of a protein even slightly will often cause shifts in the positions of backbone atoms. This is difficult to predict in the context of protein design because modeling such movement is computationally expensive and protein design applications often have to consider a very large number of potential sequences.
Therefore, most protein design applications use a fixed backbone approximation. In this case, all backbone atoms are held fixed in space and do not move in response to mutations of side-chain residues. This is not physically accurate, but has been shown to be reasonable through the success of previous designs.
Important
The biggest impact of the fixed backbone approximation is that the backbone cannot move in response to slight clashes that might occur between side-chains, when it would do so in nature. This makes the repulsive portion of most Van-der-Waals potentials overestimate the penalty for clashing sidechains. Different labs have come up with different solutions to this problem, some of which are discussed in the section about energy functions.