On Thu, Oct 16, 2008 at 4:30 PM, Tyler Reddy <TREDDY@xxxxxx> wrote:
Hi, So what I've been doing for NOE data is using the DSN command in vnmrj (which gives S/N ratio) and then converting this to the rms value by RMSD = highest peak/(S/N)
Does this give a S/N ratio for each peak? You could divide this value by the peak hight and the resultant RMSD values should be the same for all peaks. I don't know how vnmrj estimates the noise though, so I can't tell you if this value is reasonable or not. I hope it misses the solvent signal when calculating this.
If I understand what is planned for the relax program, I would want rms values for each of the T1 and T2 spectra at all field strengths, but I'm not sure if it would be worthwhile to collect error information for the different delays within a given spectrum (i.e. rms at 0.01 s and 0.05 s for a single spectrum). In principle I would think it should be comparable, and I don't want to make this unnecessarily complicated.
Actually the noise changes, usually most pronounced in the R2. I'm not sure why but I think this has something to do with the total power of the spectrum and Parseval's theorem (http://en.wikipedia.org/wiki/Parseval's_theorem). My guess would be that the total power decreases as the relaxation period increases as this is related to the intensity of the signals. Whether or not this it the reason, what you usually see is that noise decreases exponentially - although much less quickly than the signals - with an increased relaxation period. So having estimates for all spectra is useful and relax will use that in the Monte Carlo simulation error analysis. This is not necessary though, but does give you more accurate error values.
In any case, do you have an estimate for how long it will take to implement the base plane rms functionality for the curve-fitting portion of relax? If there is something I can do to help feel free to let me know. Ultimately I'd like to compare results from a few different programs, so if it will be a few weeks I might try an alternate method first and then come back and compare the results at various stages with relax. That said, I don't know if base plane rms will work from the get-go with any other packages, and a lot of the alternatives may not be user friendly.
I'm not sure, but this could be very quick or very slow depending on how complex it ends up. If lucky I can do it in a couple of hours. If unlucky it could take a couple of weeks. The infrastructure to support this is already in relax, so I would guess that it will be to the faster side of things. If it isn't fast enough for you though, you can use the other programs to get the R1 and R2 and errors and then use relax to do the model-free analysis. And if you really, really want, you could always implement this yourself. But then you'd need lots of time and have to be prepared to learn python and how open source projects work. Regards, Edward