[gmx-developers] Group cutoff-scheme simulation temperature affected by nh-chain-length
Du, Yu
duyu at sioc.ac.cn
Fri Nov 3 02:59:56 CET 2017
Hi Mark and Dear GMX-Developers,
Mark, thank you very much. I got your point. The difference is about the combination rule (5.3 and 5.4).
Geometric mean of c6 (comb-rule=1) is different from the arithmetic mean of sigma (comb-rule=2)
amber99sb-ildn is designed under comb-rule=2, althought the average difference is little.
Is there any means to adapt the source code of src/gromacs/tables/forcetable.cpp or others to get tabulized potential work with amber.ff whose comb-rule=2?
Yu
-----Original Messages-----
From:"Mark Abraham" <mark.j.abraham at gmail.com>
Sent Time:2017-11-03 09:24:19 (Friday)
To: gmx-developers at gromacs.org
Cc: GMX-developer <gromacs.org_gmx-developers at maillist.sys.kth.se>
Subject: Re: [gmx-developers] Group cutoff-scheme simulation temperature affected by nh-chain-length
Hi,
On Fri, Nov 3, 2017 at 2:12 AM Du, Yu <duyu at sioc.ac.cn> wrote:
Hi Mark and Dear GMX-Developers,
I tried the gmx check -s1 -s2, it gives comparison each recording steps.
That sounds like your comparing the trajectories. I intended you to compare the .tpr files, because the combinations are computed in that. Yours may differ.
But I need the reason of not changing comb-rule 2to1 and sigma,epsilon to c6,c12 in amber ff.
Could you explain why the combined parameters are different and whose combined parameters are different from?
See the equations in section 5.3.2 of the reference manual. Only in some cases with the parameters for a pair of atom types be the same in the two rules. Whether that applies is up to you, but if so then the gmx compare will confirm your deduction.
I have adapted the sigma and epsilon of ffnonbonded.itp to c6 and c12, comb-rule 2 to 1, I think it's reasonable and these places (forcefield.itp and ffnonbonded.itp) are where one only needs to adapt.
In so doing, you surely assumed that you were treating an A-A interaction. But the A-B interactions are the relevant ones for *combining* parameters.
But Prof. Hess said I should never change the comb-rule of ff. Thanks for any detailed information.
Because it does not work for e.g. proteins.
Mark
Best,
Yu
-----Original Messages-----
From:"Mark Abraham" <mark.j.abraham at gmail.com>
Sent Time:2017-11-02 22:52:54 (Thursday)
To:gmx-developers at gromacs.org
Cc: GMX-developer <gromacs.org_gmx-developers at maillist.sys.kth.se>
Subject: Re: [gmx-developers] Group cutoff-scheme simulation temperature affected by nh-chain-length
Hi,
The general reason is that the combined parameters are different. Whether that matters depends on your system composition. You can see if that affects you by doing gmx compare -s1 -s2
Mark
On Thu, Nov 2, 2017 at 3:42 PM Du, Yu <duyu at sioc.ac.cn> wrote:
Dear Prof. Hess,
I compared the MD results of comb-rule=1 and 2 amber99sb-ildn ff and found that there is little difference between them. The MD results of original amber ff and adapted are pasted in the end of this email.
Could you give the reason that change should never be made to the comb-rule and the coorsponding ffnonbonded.itp file?
Thanks a lot.
Regards,
Yu
##########################original amber ff############################
Statistics over 25001 steps using 251 frames
Energies (kJ/mol)
Angle Proper Dih. Improper Dih. LJ-14 Coulomb-14
4.02507e+03 5.15818e+03 2.75373e+02 2.11916e+03 1.69171e+04
LJ (SR) Coulomb (SR) Coul. recip. Potential Kinetic En.
6.20228e+04 -5.36969e+05 3.15506e+03 -4.43296e+05 8.45579e+04
Total Energy Conserved En. Temperature Pressure (bar) Constr. rmsd
-3.58738e+05 -4.39556e+05 3.00296e+02 -1.81660e+02 0.00000e+00
Total Virial (kJ/mol)
3.01440e+04 1.75821e+01 -1.58821e+01
1.84730e+01 3.00339e+04 -9.97169e+01
-1.46410e+01 -1.00240e+02 3.00330e+04
Pressure (bar)
-1.84467e+02 -3.37456e+00 2.10386e+00
-3.46046e+00 -1.75964e+02 1.04734e+01
1.98421e+00 1.05238e+01 -1.84549e+02
T-Protein T-non-Protein
2.99844e+02 3.00324e+02
M E G A - F L O P S A C C O U N T I N G
NB=Group-cutoff nonbonded kernels NxN=N-by-N cluster Verlet kernels
RF=Reaction-Field VdW=Van der Waals QSTab=quadratic-spline table
W3=SPC/TIP3p W4=TIP4p (single or pairs)
V&F=Potential and force V=Potential only F=Force only
Computing: M-Number M-Flops % Flops
-----------------------------------------------------------------------------
Pair Search distance check 14291.569978 128624.130 0.5
NxN Ewald Elec. + LJ [F] 184675.631168 12188591.657 51.1
NxN Ewald Elec. + LJ [V&F] 2250.700664 240824.971 1.0
NxN Ewald Elec. [F] 160299.489552 9778268.863 41.0
NxN Ewald Elec. [V&F] 1953.462840 164090.879 0.7
1,4 nonbonded interactions 127.655106 11488.960 0.0
Calc Weights 2540.801628 91468.859 0.4
Spread Q Bspline 54203.768064 108407.536 0.5
Gather F Bspline 54203.768064 325222.608 1.4
3D-FFT 69761.190336 558089.523 2.3
Solve PME 48.401936 3097.724 0.0
Reset In Box 42.311124 126.933 0.0
CG-CoM 42.412752 127.238 0.0
Angles 88.678547 14897.996 0.1
Propers 139.355574 31912.426 0.1
Impropers 10.650426 2215.289 0.0
Virial 10.413396 187.441 0.0
Stop-CM 8.536752 85.368 0.0
Calc-Ekin 169.413876 4574.175 0.0
Lincs 165.281870 9916.912 0.0
Lincs-Mat 3504.083832 14016.335 0.1
Constraint-V 1040.480273 8323.842 0.0
Constraint-Vir 12.528823 300.692 0.0
Settle 583.493470 188468.391 0.8
-----------------------------------------------------------------------------
Total 23873328.747 100.0
##########################original amber ff############################
##########################comb-rule=1 amber ff############################
Statistics over 25001 steps using 251 frames
Energies (kJ/mol)
Angle Proper Dih. Improper Dih. LJ-14 Coulomb-14
4.01117e+03 5.20573e+03 2.78324e+02 2.02413e+03 1.68747e+04
LJ (SR) Coulomb (SR) Coul. recip. Potential Kinetic En.
6.22595e+04 -5.37413e+05 3.13120e+03 -4.43628e+05 8.45492e+04
Total Energy Conserved En. Temperature Pressure (bar) Constr. rmsd
-3.59079e+05 -4.40129e+05 3.00265e+02 -1.82026e+02 0.00000e+00
Total Virial (kJ/mol)
3.01455e+04 -9.34378e+01 -1.54004e+02
-9.20285e+01 3.00302e+04 -4.38835e+01
-1.53975e+02 -4.42383e+01 3.00377e+04
Pressure (bar)
-1.89288e+02 8.10551e+00 1.64126e+01
7.96965e+00 -1.76137e+02 7.49214e-01
1.64098e+01 7.83424e-01 -1.80654e+02
T-Protein T-non-Protein
3.00317e+02 3.00262e+02
M E G A - F L O P S A C C O U N T I N G
NB=Group-cutoff nonbonded kernels NxN=N-by-N cluster Verlet kernels
RF=Reaction-Field VdW=Van der Waals QSTab=quadratic-spline table
W3=SPC/TIP3p W4=TIP4p (single or pairs)
V&F=Potential and force V=Potential only F=Force only
Computing: M-Number M-Flops % Flops
-----------------------------------------------------------------------------
Pair Search distance check 14608.461222 131476.151 0.6
NxN Ewald Elec. + LJ [F] 183977.095776 12142488.321 51.1
NxN Ewald Elec. + LJ [V&F] 2242.460760 239943.301 1.0
NxN Ewald Elec. [F] 159476.727584 9728080.383 40.9
NxN Ewald Elec. [V&F] 1943.414296 163246.801 0.7
1,4 nonbonded interactions 127.655106 11488.960 0.0
Calc Weights 2540.801628 91468.859 0.4
Spread Q Bspline 54203.768064 108407.536 0.5
Gather F Bspline 54203.768064 325222.608 1.4
3D-FFT 69761.190336 558089.523 2.3
Solve PME 48.401936 3097.724 0.0
Reset In Box 42.345000 127.035 0.0
CG-CoM 42.412752 127.238 0.0
Angles 88.678547 14897.996 0.1
Propers 139.355574 31912.426 0.1
Impropers 10.650426 2215.289 0.0
Virial 10.413396 187.441 0.0
Stop-CM 8.536752 85.368 0.0
Calc-Ekin 169.413876 4574.175 0.0
Lincs 168.515866 10110.952 0.0
Lincs-Mat 3570.772848 14283.091 0.1
Constraint-V 1043.984536 8351.876 0.0
Constraint-Vir 12.566057 301.585 0.0
Settle 583.673648 188526.588 0.8
-----------------------------------------------------------------------------
Total 23778711.227 100.0
##########################comb-rule=1 amber ff############################
-----Original Messages-----
From:"Du, Yu" <duyu at sioc.ac.cn>
Sent Time:2017-11-02 19:45:40 (Thursday)
To: GMX-developer <gromacs.org_gmx-developers at maillist.sys.kth.se>
Cc:
Subject: Re: [gmx-developers] Group cutoff-scheme simulation temperature affected by nh-chain-length
Thanks for Prof. Berk's prompt fix.
You advised that "You should never change the combination rule of a force field" in Bug #2286 and because tabulized potential needs comb-rule=1, what should I do to use the amber99sb-ildn ff? (ff needed for protein-ligand simulation system) Or in Gromacs, I can only use the group cutoff-scheme and tabulized potential with gromos ff?
Could you please explain the consequence I may get when I use the comb-func=1 amber99sb-ildn ff which is changed by myself?
Best,
Yu
-----Original Messages-----
From:"Du, Yu" <duyu at sioc.ac.cn>
Sent Time:2017-11-02 14:00:40 (Thursday)
To: GMX-developer <gromacs.org_gmx-developers at maillist.sys.kth.se>
Cc:
Subject: Re: [gmx-developers] Group cutoff-scheme simulation temperature affected by nh-chain-length
Dear GMX-Developers and Shirts,
To reproduce the the problem, I use the material in GROMACS Tutorials Lysozyme in Water.
Yes, Prof. Shirts. If I use standard vdwtype=cut-off, coulombtype=pme and cutoff-scheme=group, the simulation temperature accords with the ref_t.
If I use gromos ff and tabulized potential, there is no temperature problem.
Now the problem is pinned on the relation between tabulized potential and thermostat under amber99sb-ildn ff.
I filed Bug #2286 in redmine.
https://redmine.gromacs.org/issues/2286
Hope you could give some advice about this issue.
Yu
-----Original Messages-----
From:"Michael R Shirts" <Michael.Shirts at Colorado.EDU>
Sent Time:2017-10-23 19:43:38 (Monday)
To: "gmx-developers at gromacs.org" <gmx-developers at gromacs.org>, GMX-developer <gromacs.org_gmx-developers at maillist.sys.kth.se>
Cc:
Subject: Re: [gmx-developers] Group cutoff-scheme simulation temperature affected by nh-chain-length
If you can file a bug report in redmine (http://redmine.gromacs.org/) with example files, that would be useful. Before you file it can you check if problem happens if you use standard cutoffs, instead of tabulated cutoffs? I wonder if there is some sort of code path that it’s going down. I will bet that combination of inputs has only very rarely been used. Also, you said the temperature jumps. Does it come back down to 310, or does it stay there?
Generally, I would recommend using v-rescale instead of Nose-Hoover. Nose-Hoover has some known flaws; increasing the chain length is supposed to help things, but doesn’t really help that much for realistic systems, just with very simple ones.
However, any Nose-Hoover change length should in principle give the same average, and did when the code was written. If it doesn’t, the code was broken somewhere, possibly because of lack of coverage in the regression tests and something getting out of sync.
Best,
~~~~~~~~~~~~~~~~
Michael Shirts
Associate Professor
michael.shirts at colorado.edu
http://www.colorado.edu/lab/shirtsgroup/
Phone: (303) 735-7860
Office: JSCBB C123
Department of Chemical and Biological Engineering
University of Colorado Boulder
From: <gromacs.org_gmx-developers-bounces at maillist.sys.kth.se> on behalf of "Du, Yu" <duyu at sioc.ac.cn>
Reply-To: "gmx-developers at gromacs.org" <gmx-developers at gromacs.org>
Date: Monday, October 23, 2017 at 12:48 AM
To: GMX-developer <gromacs.org_gmx-developers at maillist.sys.kth.se>
Subject: [gmx-developers] Group cutoff-scheme simulation temperature affected by nh-chain-length
Dear GMX Developer and Prof. Hess,
I think this question is not common to GMX users, so I post it here.
Prof. Hess is very familiar with the tabulized potential and group cutoff scheme, so I also CC this email to you. Any advice will be appreciated.
The Backgroud:
I simulated protein-ligand system (ref_t=300) with amber99sb-ildn ff. I used the normal table6-12.xvg tabulized potential to reproduce this problem.
The Problem:
In NVT simulation, integrator=md-vv, Tcoupl=nose-hoover, coulombtype=pme-user, vdwtype=user. I found that at the start of the simulation, there is a temperature jump to around 310K (average 310K in 100ps) higher than the ref_t (300K) under nh-chain-length=10.
1)I tuned and spotted the nh-chain-length parameter. If I used nh-chain-length=1 there is no temperature jump and temperature can converge very well to 300K.
2)I replicated the same experiment in verlet scheme, the temperature converged to ref_t 300K in 100ps and no jump in nh-chain-length=2, which can 50K temperature jump (to 350K).
The Question:
1)How will the nh-chain-length parameter affect the simulation temperature under the group cut-off scheme?
2)Is it valid to use NH (nh-chain-length=1) not NH chain (nh-chain-length>1) to simulate the protein-ligand system?
Thanks for any advice.
--
Du, Yu
PhD Student,
Shanghai Institute of Organic Chemistry
345 Ling Ling Rd., Shanghai, China.
Zip: 200032, Tel: (86) 021 5492 5275
--
Gromacs Developers mailing list
* Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List before posting!
* Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
* For (un)subscribe requests visit
https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developers or send a mail to gmx-developers-request at gromacs.org.
--
Du, Yu
PhD Student,
Shanghai Institute of Organic Chemistry
345 Ling Ling Rd., Shanghai, China.
Zip: 200032, Tel: (86) 021 5492 5275
--
Gromacs Developers mailing list
* Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List before posting!
* Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
* For (un)subscribe requests visit
https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developers or send a mail to gmx-developers-request at gromacs.org.
--
Du, Yu
PhD Student,
Shanghai Institute of Organic Chemistry
345 Ling Ling Rd., Shanghai, China.
Zip: 200032, Tel: (86) 021 5492 5275
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-developers/attachments/20171103/d86ea97e/attachment-0001.html>
More information about the gromacs.org_gmx-developers
mailing list