> on a similar note, it would be useful if the overall ensemble RMSD
> (precision) could be shown somewhere (or have I just missed this
> functionality?).
This has already been added and will appear in the next release.
> I also noticed that for the per-residue RMSD both the backbone and
> all-atom (all-HEAVY-atom?) RMSDs are shown. However, for the
> per-structure RMSDs it just says "RMSD" with no indication of what atoms
> have been included.
To be clear, the backbone category is strictly heavy and backbone, i.e. N,
CO, CA and "all atom" really is all atoms.
The per-model (and per-ensemble when it arrives) RMSDs are all-atom, but
the contributions of individual atoms are variance weighted as mentioned
below.
The documentation, when complete, will describe all of this.
> Perhaps the per-structure and ensemble RMSDs could also be split into
> backbone and all-heavy-atom?
Sure, if it helps.
> Maybe longer term a rather more comprehensive RMSD calculation routine with
> greater flexibility for choosing residues/atoms could be included?
Ah. The way that I have done things is to try to eliminate any need to do
any fiddling with residue or atom selections. The superposition from the
inbuilt method should be very good indeed, and aims to give a solution
fairly close to global more probabilistic methods.
Unless there is a problem with the current method I would rather not have
to extend the GUI to allow arbitrary residue/atom comparison. Also, while
I will include more structure functionality as and when I have time, don't
forget that Analysis is primarily an NMR program.
> 1. What method is being used here?
The underlying basis of the method is singular value decomposition to
calculate an optimum rotation for two coordinate sets with co-located
centroids. Each atom is weighted, both for the SVD and centroid
calculation. There are multiple rounds of superimposition to get the
ensemble from pairwise model comparison and to better refine the weights.
Initially the weights come from the atomic masses, but in the laster
stages of the alignment the weights come from the atomic RMSDs calculated
in the earlier round. Accordingly, the dissimilar parts have
proportionately little influence on the final ensemble (and thus you don't
need to choose the 'structured' residues). Precisely, the weights become
one over the mean squared deviation (so for an atom RMSD of 0.1 Å its
weight is 100, but an RMSD of 3.0 Å its weight is 0.111.) and then in the
last stage exp(-(RMSD/k)^2), which is even more strict (with k=0.8, 0.1 Å
gives weight=0.98, 3.0 Å gives weight=0.00000078),
The procedure is:
Co-locate (weighted) centroids
SVD superimpose all models to first model
Calculate mean and closest model to mean
Re-calculate weights based upon RMSD
SVD superimpose again with new weights to closest to mean
Repeat once again using RMSD based weights rather than atomic weights
Convergence is pretty good. And if you are in any doubt you can press the
"Align" button again.
> 2. Are the stored coordinates of the "aligned" structures then
> transformed into their superposed values, and is that persistent?
Yes and yes.
> 3. Does the alignment functionality deserve a button of its own?
I don't think there is any harm in combining a persistent superposition
with the RMSD calculation, so I'd like to keep it coupled. I guess the
improved documentation should help.
> 4. If the answer to 1 is not maximum likelihood (like theseus), is
> there scope for restricting which residues are used in the aligment?
Hopefully this is now answered.
T.
-------------------------------------------------------------------------------
Dr Tim Stevens Email: [log in to unmask]
Department of Biochemistry [log in to unmask]
University of Cambridge Phone: +44 1223 766018 (office)
80 Tennis Court Road +44 7816 338275 (mobile)
Old Addenbrooke's Site +44 1223 364613 (home)
Cambridge CB2 1GA WWWeb: http://www.bio.cam.ac.uk/~tjs23
United Kingdom http://www.ccpn.ac.uk
-------------------------------------------------------------------------------
------ +NH3CH(CH(CH3)OH)C(O)NHCH(CH(CH3)CH2CH3)C(O)NHCH(CH2CH2SCH3)CO2- -------
-------------------------------------------------------------------------------
|