Hi Edward,
Yes, relax could become one of the most complete programs, and this
would be a strong point for it. Concerning features that would make
relax really unique, I have a suggestion.
It would be great if relax allowed to perform simultaneous fits of
multiple data sets. For example, in a number of cases people may have
CPMG relaxation dispersion on methyl 13C, and backbone (15N or CA).
Likewise, fitting SQ and MQ at the same time would be another instance
for such simultaneous fits.
One might also think about situations where R1rho and CPMG dispersions
might be fit at the same time (although this is a bit more special, as
it requires that the time scale can be detected by both methods).
In all these cases, one would like to have a set of common fit
parameters. Typically, the population and the kinetics would be such
common parameters. Then, one would have fit parameters that are specific
to each data set, e.g. the chemical shift difference. There might also
be cases where e.g. the 13C chemical shift difference is fit from both
SQ and MQ dispersions, but the MQ has, in addition, also a contribution
from the 1H shift difference.
Then, it would of course also be great if such a fit could be done
simultaneously for a set of residues.
In principle, implementing such joint fits should be rather
straightforward, as the target function would contain all these data
sets from different residues at the same time, and the fit parameters
would contain all the various rates, populations and shift differences.
Of course, the number of fit parameters quickly increases (e.g. the
fitted plateau values and the chemical shift differences are separate
for each data set and residue). So there will be some careful thinking
involved when designing the minimization procedure.
Do you see ways to do this? Another challenge will be that this is
flexible enough such that the user can chose which data sets to fit, and
which parameters affect which data set - such as: fit 13C SQ, 13C MQ,
15N SQ CPMGs for residues X, Y, and Z, and optimize parameters XX, YY,
ZZ,... Possibly, if one could then even fix some of the parameter, such
that they are not fitted, this could be helpful. For example, one might
know the chemical shift difference for some of the nuclei from somewhere...
Any thoughts on this?
Paul
On 25.07.13 10:43, Edward d'Auvergne wrote:
Hi,
The MQ dispersion analysis is also most welcome to be added to relax.
For the paper I'm thinking of creating a table of all the CPMG-type
analytic and numeric models, all R1rho-type analytic and numeric
models, and maybe now also the MQ models. These would be the rows,
and the columns would be the published software. Then there would be
ticks for which software supports which models. It should make a
convincing argument for using relax. Do you know of any published
software for multi-quantum dispersion data analysis?
Cheers,
Edward