Hi,
I'm guessing you're talking about the dispersion auto-analysis, right?
Turning R20 off in the grid search is trivial when not using the
auto-analysis. Though using the minimum R2eff value for an initial
R20 value is not implemented yet - though a new user function could be
created for this. Then the auto-analysis could have a new argument
added to turn this feature on (it will be off by default). This will
cause the new user function to be called and the R20 values not
minimised in the grid search. Such an idea could be implemented in
parts - the user function with system tests first, then tackle the
auto-analysis. Would this be of benefit for your analyses? Would you
like to give it a go?
There is only one disadvantage and that is if there is still a long
way from the lowest R2eff value and R20. The reason is because R20
dominates the relaxation equations. It acts as a scaling factor as
well, essentially removing all non-dispersion contributions to R2
prior to the fitting of dispersion parameters. Therefore a small
change in R20 will result in a large change of the other dispersion
parameters. So there is a significant risk that a bad value of R20
will cause the other parameters to be way, way off. This is similar
to the model-free analysis where the diffusion tensor contributes up
to 80% of the relaxation and the internal motions the small amount
remaining. So if you get the diffusion tensor wrong, you end up with
significant artificial motions everywhere. That being said, a course
grid search will have exactly the same problem. From previous posts
I'm guessing you've already seen this.
Regards,
Edward
On 24 April 2014 17:05, Troels Emtekær Linnet <tlinnet@xxxxxxxxxxxxx> wrote:
Hi Edward.
Could the grid search be made not to include R20?
For cpmg experiments, the best guess would be R2eff at the highest cpmg
freq.
For a 2 field cpmg for CR72, that would drop gridNr^5 to gridNr^3.
For standard 21 grid, it would make it 441 times faster.
Best
Troels
_______________________________________________
relax (http://www.nmr-relax.com)
This is the relax-devel mailing list
relax-devel@xxxxxxx
To unsubscribe from this list, get a password
reminder, or change your subscription options,
visit the list information page at
https://mail.gna.org/listinfo/relax-devel
Re: Speed up of grid search