Hi, This proposal could be quite useful, and simple to implement in the full_analysis.py script. I could add a variable called something like EXCLUDE which points to another file. This would then be used with the spin system deselection user function (unselect.read() in the 1.2 line and deselect.read() in the 1.3 line). These would then not only be excluded in the multi_model() method of the script, but also in the 'init' round of optimisation to cover all places in the script. Then you could take any relax model-free results file as the input, and the spins in EXCLUDE will never be used in the analysis. Is this along the lines of what you were thinking? The UNRES file would still be there, but this would serve a different purpose than then EXCLUDE file. Regards, Edward On Tue, May 27, 2008 at 3:55 PM, Sébastien Morin <sebastien.morin.1@xxxxxxxxx> wrote:
Hi, First, thanks for answering this question. Here are some precisions... If one wants to analyze a protein and doesn't know much about the global dynamics, the first step will probably be to make a full analysis on all of the protein. Then, if that person thinks an hybrid would be better, tests could be made excluding, let's say, the flexible C-terminus. To save time, one could take advantage of the already minimized "local_tm" run... Hence, this user could just exclude the C-terminus residues with the 'UNRES' variable in the different scripts for the other different diffusion tensor (but not the "local_tm" which is already available, even though for more residues). So, my question was about knowing if initializing a diffusion tensor with more residues would be bad. You answered yes, i.e. that one should initialize diffusion tensors using only the wanted residues. So, would it be a good idea to add a variable in the full_analysis script so a different 'UNRES' variable can be used for "local_tm" and other diffusion tensors..? If a different 'UNRES' variable would be used, the "init" run could only use the selected residues and exclude others, even if they are present in the "local_tm" run... This would allow one to skip the "local_tm" run when trying some hybrids... This would save lots of time... What do you think of this proposal ? Is this useful ? Is this feasible ? Thanks ! Séb Edward d'Auvergne wrote:On Sun, May 11, 2008 at 10:33 PM, Sébastien Morin <sebastien.morin.1@xxxxxxxxx> wrote:Hi, When using the full_analysis.py script, unselecting residues is possible by inputing a list of residues under the variable name UNRES. Two approaches are possible. 1. Excluding residues a priori. 2. Excluding residues after an initial test where at least the local_tm run as been completed. Let's say, for example, that one wants to produce an hybrid after an initial test with all residues fitted together.Both are fine because the local tm models are completely independent of each other.In situation 2, one may want to re-use the local_tm run and exclude residues from there on (to save time). This works fine for subsequent rounds for diffusion tensors sphere, prolate, oblate and ellipsoid. However, the "init" round (before round 1) still includes the residues to be excluded, since those were present in the local_tm run. Is this a problem ? Will this presence of subsequently excluded residues bias the diffusion tensor and avoid the user from finding the global minimum ?This is a problem as the initial diffusion tensor optimisation should only include spins located within this isolated diffusion unit.Should excluded residues be excluded from ALL the runs, including the local_tm run (which serves for creating the "init" round of other runs) ?Well, that depends on why you are excluding them. If the relaxation data is rubbish, then these should be excluded everywhere. This is the case with the spin list in UNRES, and this will be propagated in full_analysis.py into the 'init' round of optimisation because of the loading of the local tm results file. For a hybrid model, this needs to be constructed differently. If you place the spins to exclude into the UNRES file, then these will be excluded in all parts of the full_analysis.py script. I hope this helps. Regards, Edward