On Thu, 2006-10-05 at 03:11 +1000, Edward d'Auvergne wrote: > Hi, > > That is good question. I have to warn you though that my opinion is > going to be very heavily biased! Essentially the way that model-free > analysis has been implemented over the last 17 years or so (since the > publication of Kay et al., 1989) is as follows: > > 1. Estimate the Brownian rotational diffusion tensor. > 2. Hold the diffusion tensor fixed and optimise each model-free model. > 3. Model-free model selection (in my opinion this is best done using > AIC model selection ;). > 4. Optimisation of the diffusion tensor parameters together with the > parameters of the selected model-free model. > 5. Repeat the steps, using the final optimised diffusion tensor as > the starting point of the next iteration, until 'convergence'. > > On top of this I have recently proposed an additional step prior to > 'model-free model selection' called 'model-free model elimination' to > remove failed model-free models. The most common way of carrying out > step 1 is to use the R2/R1 ratio (Kay et al., 1989). relax can not > only implement this data analysis chain but, due to it's modularity > and flexibility, it can also implement many of the different published > variations to this approach. > > There is a sample script called 'full_analysis.py' distributed with > relax which implements a radically different approach to Kay's > paradigm. Rather than starting with the diffusion tensor and ending > with the model-free parameters, this new model-free optimisation > protocol applies this logic in reverse. It starts by optimising the > model-free models and finishes by optimising the diffusion tensor. > The benefits of this approach is that it avoids the pitfalls of > obtaining the initial diffusion tensor estimate, avoids the hidden > motion problem (Orekhov et al., 1995, Orekhov et al., 1999a, Orekhov > et al., 1999b), and avoids under-fitting (which causes artificial > motions to appear). > > I presented this new protocol on a poster at the 2006 ICMRBS > conference in Goettingen and I currently have a number of submitted > manuscripts which, unfortunately, are not published yet. These papers > will demonstrate the application and performance of the new model-free > optimisation protocol. However all the steps of the analysis are > described in fine detail at the start of the 'full_analysis.py' > script. > > Sorry about all that biased, unpublished opinion. In summary relax > can be used to implement most of the data analysis protocols in the > literature. I hope that answers your question. > > Edward > > > References: > Kay, L. E., Torchia, D. A., and Bax, A. (1989) Biochem. 28(23), 8972-8979. > Orekhov, V. Y., Korzhnev, D. M., Diercks, T., Kessler, H., and > Arseniev, A. S. (1999a) J. Biomol. NMR 14(4), 345-356. > Orekhov, V. Y., Korzhnev, D. M., Pervushin, K. V., Hoffmann, E., and > Arseniev, A. S. (1999b) J. Biomol. Struct. Dyn. 17(1), 157-174. > Orekhov, V. Y., Pervushin, K. V., Korzhnev, D. M., and Arseniev, A. S. > (1995) J. Biomol. NMR 17(1), 157-174. > >
I can add my observations, from spending much time hitting my head against these problems in a few different systems. I find that Lewis Kay's approach and Edward's new proposal (the 'full_analysis.py' approach) work well for differing systems. For very good data, both work well and converge on the same result. For data has a few 'outliers' (eg. residues affected by Rex or complex ns motions) Kay's approach can fail because the initial estimate of the diffusion constant is not good enough. On the other hand, Edward's approach seems more robust to these types outliers, but in my hands sometimes requires a very precise data set in order to converge. One clear benefit of Edward's approach is that failure is obvious - it simply doesn't converge on a diffusion ternsor, and chi2 eventually starts to rise as the tensor drifts into the wrong area. On the other hand, judging the success of the Kay approach can be much more difficult. In conclusion, my advice is to try both approaches on a carefully pruned dataset - it is well worth spending time on getting this right first up, because everything else depends on it. Hope this helps, Chris _______________________________________________ relax (http://nmr-relax.com) This is the relax-users mailing list [email protected] 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-users
