rmsavgcorr

Calculate RMS average correlation curve for a COORDS data set.

rmsavgcorr [crdset <crd set>] [<name>] [<mask>] [out <filename>] [mass] [stop <maxwindow>] [offset <offset>] {reference <ref file> parm <parmfile> | first}

[crdset <crd set>] COORDS data set to use (if not specified the default COORDS set will be used).
[<name>] Output data set name.
[<mask>] Atoms to calculate RMS average correlation for.
[out <filename>] Output filename.
[mass] Mass weight the RMSD calculation.
[stop <maxwindow>] Only calculate RMS average correlation up to <maxwindow>.
[offset <offset>] Skip every <offset> windows in calculation.
[first] Use first averaged frame as reference for each window (default).
[reference <ref file> [parm <parmfile>] Use reference file (with specified parm) as reference for each window.

The RMS average correlation (more details in this reference, RAC) is calculated as the average RMSD of running-averaged coordinates over increasing window sizes (or lag).

Output has format:
<WindowSize> <RAC>
The first entry has a window size of 1, and so is just the average RMSD of all frames to the specified reference structure. The second entry has a window size of two, so it is the average RMSD of all frames averaged over two adjacent windows to the specified reference, and so on. The RAC will be calculated up to the number of frames minus 1 or the value specified by stop, whichever is lower. The offset can be used to speed up the calculation by skipping window sizes. To calculate mass-weighted RMSD specify mass. Note that to reduce memory costs it can be useful to strip all coordinates not involved in the RMS fit from the system prior to specifying ’rmsavgcorr’.

For example, to calculate the correlation of C-alpha RMSD of residues 2 to 12:

strip !(:2-12@CA)
rmsavgcorr out rmscorr.dat

The curve generated by RAC decays towards zero due to the way RAC is defined. By the time the “lag” is N-1 (where N is the total number of frames) you have only two averaged coordinates: call them Avg1 (averaged over 1 though N-1 frames) and Avg2 (averaged over 2 through N frames). Barring any extraordinary circumstances the RMSD between Avg1 and Avg2 will almost certainly be quite low.

The RAC is a way to probe the time scales of interesting events. Any deviation from a smoothly decaying curve is an indication that there are some significant structural differences occurring over that time interval. RAC curves can be particularly useful when comparing independent simulations of the same system. One thing to keep in mind that since the underlying metric is RMSD, it can be sensitive to the reference frame you choose. It may be useful to try looking at both RAC from the first frame, as well as an averaged reference frame.

For an example of use see Galindo-Murillo et al., in particular Figure 2.