Hello Romel-- > > I used an adapted version of saxs_EI/dock.py to setup an run with RDC and SAXS > data. I am struggling to get the desired results. I have only recently started > using xplor-nih and I am not sure whether my script is sound. > > Basically, what I would like to achieve is to calculate structures of a two > domain protein, whereby the domain-domain orientations are defined by RDCs and > the overall shape information derived from SAXS. I should mention that the > inter-domain linker comprises 180 residues and there are barely any > experimental > constraints for the linker except for a few RDCs (and SAXS).
That linker may cause problems, but first you need to ascertain that the RDCs from each domain are consistent with a common alignment tensor. > > Crystal structures of the two domains have been solved previously - I have > made > use of them and modelled the inter-domain linker into one pdb to create a > tandem > domain construct. Briefly, the N-, linker and C-terminal domains comprise of > residues 40:168, 169:348 and 349:460 respectively. > > I have stumbled across saxs_EI/dock.py and believed it to be a good starting > point for my calculations. The flow I am trying to follow is described below > and > the script is attached at the end of this message: > > - Randomize N-terminal domain position (keep C-terminal domain > fixed). > > - Calculate C-terminal alignment tensor and keep it fixed. > > - Rotate N-terminal domain until the RDCs are satisfied. > > - Keep domains fixed; only allow translation of the domains. > > - High-temperature dynamics on the linker (treat domains as rigid > bodies) > > - Refine structure using SAXS. > > Is there a better way to do this? Or a better template than saxs_EI/dock.py? > > When I run dock.py, I have to set cutoffRfactor at a rather ‘high’ value (>32) > for the simulation to pass the RDC step but then I end up with a structure for > the linker that is not meaningful, i.e., numerous ‘unusual bonds’, steric > clashes, etc. The domains haven’t changed during the simulation, as > anticipated, > but they’re not aligned according to the RDCs as I would have hoped. It also > appears as though the structure is not refined against the SAXS data too well. > > In the case where cutoffRfactor is set at a more reasonable value of 21, the > simulation is stuck in an endless cycle trying to find the ‘right’ tensor. > The cutoff R-factor is determined as the weighted average of the R-factors from fits to the two domains- you would add a few percent to this value. I presume that you have not determined any contact information from chemical shift perturbation. This would be quite useful. Looking at your script, I don't see anything obviously wrong, but your long linker may require a different treatment. One approach might be to remove it altogether, but that will likely mess up the saxs calculation, since it is such a large fraction of the structure. Please verify that the RDCs are consistent with a single structure, and that you do not have chemical shift mapping information, and we'll think of other approaches. best regards-- Charles _______________________________________________ Xplor-nih mailing list [email protected] https://dcb.cit.nih.gov/mailman/listinfo/xplor-nih
