Hi Edward. Can you point to any reference, where kex is measured at higher values?
I see no logic in having linear constraints for a search space higher than physically possible? The edge cases you mention (https://gna.org/bugs/?22024) is a very good example. I later learned for this case, that data was acquired at 3 different temperatures, to allow for robust interpretation of the data. Best Troels 2014-08-19 10:58 GMT+02:00 Edward d'Auvergne <[email protected]>: > Hi Troels, > > I am unable to download that reference and the book with the reference > in Goettingen is 500 km away from where I am now. Would you be able > to send me a copy in a private mail? Cheers. > > I'm not convinced that this change should be made. The constraint of > 1e6 was simply a large upper bound whereby no exchange should be > visible. The constraint is used to prevent kex from heading out to > infinity. So changing it to 1e4 or 1e5 will have no effect on this > purpose of the constraint. > > What worries me here is that although exchange is very close to zero > at these new limits, they are not exactly zero. This is especially > the case for data with very, very small errors (test data with zero > errors will show this). So the edge cases you have mapped out with > OpenDX (https://gna.org/bugs/?22024 for example) will be highly > effected by such a change. But this would be an edge case on the > other side of the space. The new limits could chop a very weak > minimum out of the optimisation space - i.e. you've put up a wall in > front of these weak minima. This is a very weak edge case, but it > nevertheless exists. As the purpose of the constraint is simply to > stop kex from going to infinity, loosing these special edge case > minima is not worth it. I would suggest reverting this change. > > Cheers, > > Edward > > > > On 15 August 2014 15:11, <[email protected]> wrote: >> Author: tlinnet >> Date: Fri Aug 15 15:11:00 2014 >> New Revision: 25024 >> >> URL: http://svn.gna.org/viewcvs/relax?rev=25024&view=rev >> Log: >> Modified the Linear Constraints for the exchange rates. >> >> For CPMG, the maximum kex should be 10^4, and for R1rho it should be 10^5. >> This is altered from the value of 10^6. >> >> The suggested restraints for 'kex' follows from article, on page 224: >> >> Nuclear Magnetic Resonance Methods for Quantifying >> Microsecond-to-Millisecond Motions in Biological Macromolecules. >> Palmer-III, Arthur G., Kroenke, Christopher D., Loria, J. Patrick >> Nucl. Magn. Reson. Biol. Macromol. B, 2001, Vol: 339, pages 204-238. >> U{DOI: >> 10.1016/S0076-6879(01)39315-1<http://dx.doi.org/10.1016/S0076-6879%2801%2939315-1>} >> >> Modified: >> trunk/specific_analyses/relax_disp/parameters.py >> >> Modified: trunk/specific_analyses/relax_disp/parameters.py >> URL: >> http://svn.gna.org/viewcvs/relax/trunk/specific_analyses/relax_disp/parameters.py?rev=25024&r1=25023&r2=25024&view=diff >> ============================================================================== >> --- trunk/specific_analyses/relax_disp/parameters.py (original) >> +++ trunk/specific_analyses/relax_disp/parameters.py Fri Aug 15 15:11:00 >> 2014 >> @@ -34,7 +34,7 @@ >> from pipe_control import pipes >> from pipe_control.mol_res_spin import exists_mol_res_spin_data, return_spin >> from specific_analyses.relax_disp.data import count_spins, >> generate_r20_key, has_exponential_exp_type, loop_cluster, loop_exp_frq >> -from specific_analyses.relax_disp.variables import MODEL_LIST_MMQ, >> MODEL_M61B, MODEL_NS_MMQ_3SITE, MODEL_NS_MMQ_3SITE_LINEAR, >> MODEL_NS_R1RHO_3SITE, MODEL_NS_R1RHO_3SITE_LINEAR, PARAMS_R20 >> +from specific_analyses.relax_disp.variables import >> MODEL_LIST_ANALYTIC_R1RHO, MODEL_LIST_CPMG_ONLY, MODEL_LIST_MMQ, >> MODEL_LIST_NUMERIC_R1RHO, MODEL_M61B, MODEL_NS_MMQ_3SITE, >> MODEL_NS_MMQ_3SITE_LINEAR, MODEL_NS_R1RHO_3SITE, >> MODEL_NS_R1RHO_3SITE_LINEAR, PARAMS_R20 >> >> >> def assemble_param_vector(spins=None, key=None, sim_index=None): >> @@ -435,6 +435,12 @@ >> def linear_constraints(spins=None, scaling_matrix=None): >> """Set up the relaxation dispersion curve fitting linear constraint >> matrices A and b. >> >> + The suggested restraints for 'kex' follows from article, on page 224: >> + Nuclear Magnetic Resonance Methods for Quantifying >> Microsecond-to-Millisecond Motions in Biological Macromolecules. >> + Palmer-III, Arthur G., Kroenke, Christopher D., Loria, J. Patrick >> + Nucl. Magn. Reson. Biol. Macromol. B, 2001, Vol: 339, pages 204-238. >> + U{DOI: >> 10.1016/S0076-6879(01)39315-1<http://dx.doi.org/10.1016/S0076-6879%2801%2939315-1>}. >> + >> Standard notation >> ================= >> >> @@ -453,12 +459,12 @@ >> phi_ex_C >= 0 >> padw2 >= 0 >> dw >= 0 >> - 0 <= kex <= 2e6 >> - 0 <= k_AB <= 2e6 >> - 0 <= kB <= 2e6 >> - 0 <= kC <= 2e6 >> + 0 <= kex <= 1e4, for CPMG >> + 0 <= kex <= 1e5, for R1rho >> + 0 <= k_AB <= 1e4 >> + 0 <= kB <= 1e4 >> + 0 <= kC <= 1e4 >> tex >= 0 >> - k_AB >= 0 >> >> >> Matrix notation >> @@ -502,19 +508,22 @@ >> | | | | | | >> | 1 0 0 | | kex | | 0 | >> | | | | | | >> - |-1 0 0 | | kex | | -2e6 | >> + |-1 0 0 | | kex | |-1e4/-1e5| >> + | | | | | | >> + | 1 0 0 | | k_AB | | 0 | >> + | | | | | | >> + |-1 0 0 | | k_AB | | -1e4 | >> | | | | | | >> | 1 0 0 | | kB | | 0 | >> | | | | | | >> - |-1 0 0 | | kB | | -2e6 | >> + |-1 0 0 | | kB | | -1e4 | >> | | | | | | >> | 1 0 0 | | kC | | 0 | >> | | | | | | >> - |-1 0 0 | | kC | | -2e6 | >> + |-1 0 0 | | kC | | -1e4 | >> | | | | | | >> | 1 0 0 | | tex | | 0 | >> | | | | | | >> - | 1 0 0 | | k_AB | | 0 | >> >> >> @keyword spins: The list of spin data containers for the >> block. >> @@ -628,14 +637,21 @@ >> j += 1 >> break >> >> - # Exchange rates and times (0 <= k <= 2e6). >> + # Exchange rates and times (0 <= k <= 1e4) for CPMG and (0 <= k <= >> 1e5) for R1rho. >> elif param_name in ['kex', 'kex_AB', 'kex_AC', 'kex_BC', 'k_AB', >> 'kB', 'kC']: >> A.append(zero_array * 0.0) >> A.append(zero_array * 0.0) >> A[j][param_index] = 1.0 >> A[j+1][param_index] = -1.0 >> b.append(0.0) >> - b.append(-2e6 / scaling_matrix[param_index, param_index]) >> + # For CPMG experiments, (0 <= k <= 1e4). >> + if spins[0].model in MODEL_LIST_CPMG_ONLY + MODEL_LIST_MMQ: >> + b.append(-1e4 / scaling_matrix[param_index, param_index]) >> + # For R1rho experiments, (0 <= k <= 1e5). >> + elif spins[0].model in MODEL_LIST_ANALYTIC_R1RHO + >> MODEL_LIST_NUMERIC_R1RHO: >> + b.append(-1e5 / scaling_matrix[param_index, param_index]) >> + else: >> + b.append(-2e6 / scaling_matrix[param_index, param_index]) >> j += 2 >> >> # Exchange times (tex >= 0). >> >> >> _______________________________________________ >> relax (http://www.nmr-relax.com) >> >> This is the relax-commits 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-commits > > _______________________________________________ > relax (http://www.nmr-relax.com) > > This is the relax-devel 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-devel _______________________________________________ relax (http://www.nmr-relax.com) This is the relax-devel 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-devel
