Re: relax use -- November 01, 2012 - 17:46

Hi Ravi,

Please see below:

On 1 November 2012 16:43, Ravi Pratap Barnwal <ravi13pb@xxxxxxxxx> wrote:

Hello,
I dont see this error when i start he relax. I do solved the problem. It seems some formatting error associated with the R1/R2 and noe files, was a culprit.

Ok, this is good to know.

 

Also, when i was checking my results from the previous run, i noticed relax did not run "prolate spheroid" and "oblate spheroid" modes. It did create a final folder with all the files. Also, the correlation time given by relax seems to be double of what we got from model-free, how can this be possible? It was based on sphere mode.

I would not worry about results from previous runs.  I would re-run everything from scratch with the formatting errors fixed, and then see.

Note that when you compare, you are comparing different protocols here.  You can set up relax to obtain almost identical results to Art Palmer's Modelfree4 program (if you switch to scripting and you really know what you are doing).  I'm assuming you mean this software.  And if you can handle FORTRAN code and change some hard-coded optimisation precision variables in Modelfree4 to the higher precision values that relax defaults to, and then implement the model-free protocol I came up with (see http://www.nmr-relax.com/refs.html#dAuvergneGooley07 and http://www.nmr-relax.com/refs.html#dAuvergneGooley08b), you can use Modelfree4 and obtain the same results as relax.  If you are using both programs as blackboxes and want to compare results, I would strongly suggest to read about and understand the fundamental concepts and protocols behind these blackboxes.  Then you should be able to understand why both programs can obtain both sets of results, and why both sets of results are different.

I also have a few questions about how you used Modelfree4.  How did you obtain your results from Modelfree4?  Which protocol did you implement and how?  How did you obtain your initial tensor estimate?  Do you know why one of the protocols you used requires an initial diffusion tensor estimate and the other does not?  How many iterations of model-free model optimisation, model selection, and diffusion tensor optimisation were required for each of the different diffusion models in Modelfree4?  And how does this compare to the number of iterations required by relax?  What happens if you use the tensor from relax as the input for Modelfree4?  Does it return to the smaller tensor?  And how does the total chi-squared value change, or how does it compare to that of relax?  Also, how did you select between the diffusion models from Modelfree4?  Maybe everything will change once you re-run your analysis in relax anyway.  I hope some of this helped.

Regards,

Edward