From Jacobson Lab Wiki

Jump to: navigation, search

Helix is a top level plop command used to predict the location of helical secondary structure.

Helix prediction in PLOP is designed to systematically explore the conformational space of a single alpha-helix in the protein (within an adjustable cutoff of the extent of movement) and predict its most energetically favorable position and orientation.

The basic syntax for the helix prediction command looks like:

 helix auto[move] [residues defining loop-helix-loop region] & 
   roll real real real
   grid real  &
   ofac real&
   deviation real &
   deltan real&
   deltac real&
   write_pdb yes/no

[edit] Specifying which helix to adjust

It is mandatory that the starting/ending residue of the helix, as well as the two flanking loops on both sides of the helix, be specified right after the “helix auto[move]” command.

The two flanking loops must be precisely specified since they are the key factors in determining the constraints on the movement of the helix, which is treated as a rigid body during the sampling. Basically, it assumes the following syntax:

 helix auto  loop_start   helix_start  helix_end loop_end


 helix auto A:20  A:25  A:39 A:46

specifies that the helix spanning from residue A:25 to A:39 are to be predicted, with the flanking loops ranging from A:20 to A:24 and A:40 to A:46 on both sides respectively as constraints on the helix movement during sampling. Typically, the specified length of the flanking loops should not exceed 8 residues (even if the loops are actually longer than that), to avoid greatly increasing the computational expense.

[edit] Algorithm and Options

As a preliminary step, the helix sampling algorithm sets up two bounding spheres centered on the two end points of the helix and creates a network of grid points within them. The grid size is given by the grid parameter, the default of which is 1.0 (Angstrom). Each grid pair on the two ends will then determine a potential central axis for the helix. That can lead to a vast number of potential helix positions. The parameter deviation can then be used to constrain the most dramatic extent of the helix movement, i.e., it will be the maximum distance allowed for either end points to deviate from its initial position. Alternatively, one can also specify the maximum deviations for the N- and C- terminus separately, using the parameters deltan and/or deltac respectively.

Since the helix end points can only uniquely determine 5 out of the 6 degrees of freedom of the helix position, a 6th parameter, which determines the angle to be rotated around the helix axis, must be specified. This is reflected in the roll parameters. Three values must be specified here: the minimum, maximum, and grid size (stepwise increment) of the rotation angle. For example:

 roll 0 360 30

means sampling the helix starting from 0 (no rotation) to 360 degrees, with a grid size of 30 degrees (along with the other 5 spatial constraints). In order to sample the relevant conformations of the helix without rotating the helix a full 360 degrees, one can also specify negative degree values. For example:

 roll -60 60 30

will explore 120 degrees of rotation, and, unlike the previous notation, will allow exploration closer to the native roll state without having to rotate a full 300 degrees to obtain the -60 degree state. The following input, however, will not work as PLOP will believe that it has a negative number of helix conformations to sample and the algorithm will crash.

 roll 300 60 30

After the geometric filtering, typically one would get a significant reduction in the number of potential helix positions under consideration. One can either explicitly write out each of these (partial) conformations to the pdb file or just suppress them, through the specification of the write_pdb option (it is usually recommended to turn this option off). The rest of the helix positions are then clustered using a K-means clustering algorithm similar to that used in the loop prediction. For this limited set of helix positions, side chains as well as the two flanking loops are added on and the whole loop-helix-loop region (LHL) is ranked ordered by the all-atom energy calculation. The lowest energy representative among all clusters is chosen as the final predicted conformation. Nevertheless, since the loop closure is done through only 1 iteration, one might wish to continue running further loop refinement to get further improvement on the prediction quality on the LHL region.

Personal tools