From: Peter Freddolino (petefred_at_ks.uiuc.edu)
Date: Wed Dec 16 2009 - 15:46:39 CST
Hi Francesco,
in order to get steepest descent-ish minimization I believe that you
have to do
run 1000
instead of minimize after turning on velocityQuenching. However,
going back to your initial questions, namd shouldn't choke on this
input. Your initial energies are still *huge*, and if you've eliminated
the possibility that they're from bad protein-protein interactions,
something about your solvation is still most likely. Have you tried
looking at both the initial system and the system right before the crash
in VMD with periodic images turned on? Also, you might want to use
namdenergy to look at the interaction energies for protein, water, and
protein-water interactions to find where the huge vdw energy is coming
from (my money is on protein-water interactions, which would mean
thinking more about your solvation parameters). Keep in mind that you
can't use the "protein" and "water" keywords with a cg model.
Best,
Peter
Francesco Pietra wrote:
> Attempts at minimizing under "steepest descent" ended in a crash. I
> wonder whether a suggestion may arise from the following.
>
> Modifications to the conf file:
>
> set temperature 0
>
> temperature $temperature
>
> langevin off
>
> # Minimization
> if {1} {
> velocityQuenching on
> maximumMove 1.5
> minimize 1000
> reinitvels $temperature
> }
>
> numsteps 10
>
> The log said:
> ............
> Info: Startup phase 7 took 0.0559781 s, 11.754 MB of memory in use
> Info: Startup phase 8 took 0.000222921 s, 11.8803 MB of memory in use
> Info: Finished startup at 15 s, 11.8803 MB of memory in use
>
> TCL: Minimizing for 1000 steps
> ------------- Processor 2 Exiting: Caught Signal ------------
> Signal: segmentation violation
> Suggestion: Try running with '++debug', or linking with '-memory paranoid'.
> [2] Stack Traceback:
> [0] /lib/libc.so.6 [0x7f4c538daf60]
> [1] _Z24sortEntries_mergeSort_v2RP12__sort_entryS1_i+0xba [0x5ce526]
> [2] _ZN20ComputeNonbondedUtil32calc_pair_energy_merge_fullelectEP9nonbonded+0x351a
> [0x592c2a]
> [3] _ZN20ComputeNonbondedPair7doForceEPP8CompAtomPP11CompAtomExtPP7Results+0xae4
> [0x57b04e]
> [4] _ZN16ComputePatchPair6doWorkEv+0xa7 [0x6e21a3]
> [5] _ZN11WorkDistrib12enqueueWorkAEP12LocalWorkMsg+0x16 [0x92ed3c]
> [6] _ZN19CkIndex_WorkDistrib31_call_enqueueWorkA_LocalWorkMsgEPvP11WorkDistrib+0xf
> [0x92ed23]
> [7] CkDeliverMessageFree+0x21 [0x9c2d71]
> [8] _Z15_processHandlerPvP11CkCoreState+0x509 [0x9c2365]
> [9] CsdScheduleForever+0xa5 [0xa4bcf5]
> [10] CsdScheduler+0x1c [0xa4b8f6]
> [11] _Z11master_initiPPc+0x280 [0x508ea0]
> [12] _ZN7BackEnd4initEiPPc+0x31 [0x508c19]
> [13] main+0x2f [0x5045af]
> [14] __libc_start_main+0xe6 [0x7f4c538c71a6]
> [15] _ZNSt8ios_base4InitD1Ev+0x52 [0x4ff9ea]
> Fatal error on PE 2> segmentation violation
>
>
> Thanks
> francesco
>
> ---------- Forwarded message ----------
> From: Francesco Pietra <chiendarret_at_gmail.com>
> Date: Wed, Dec 16, 2009 at 9:13 AM
> Subject: Re: vmd-l: Re: namd-l: Fwd: conf file for coarse grained simulation
> To: Axel Kohlmeyer <akohlmey_at_gmail.com>
> Cc: NAMD <namd-l_at_ks.uiuc.edu>, Peter Freddolino <petefred_at_ks.uiuc.edu>
>
>
> Thanks a lot for what you wrote. Solvation from scratch with Boundary
> 4.0 and Padding 19 (15+4) did not help, or not fully. Minimization at
> const. volume (with protein+bilayer restrained) halted at step 532,
> out of 1000 set steps.
>
> The reason to posting now, before I continue to search the right
> avenue, is that now the VDW energy has decreased conspicuously, from
> initial
>
> MINIMIZER SLOWLY MOVING ATOMS WITH BAD CONTACTS DOWNHILL
> PRESSURE: 3 1.5157e+10 1.03589e+10 4.97618e+09 9.72122e+08 1.2445e+10
> -5.9558e+08 -2.86152e+09 -1.00705e+10 1.38611e+09
> GPRESSURE: 3 1.5157e+10 1.03589e+10 4.97618e+09 9.72122e+08 1.2445e+10
> -5.9558e+08 -2.86152e+09 -1.00705e+10 1.38611e+09
> ENERGY: 3 1578.4243 3635.7565 1250.6257
> 0.0000 -5656.7864 9999999999.9999 0.0000
> 0.0000 0.0000 9999999999.9999 0.0000
> 9999999999.9999 9999999999.9999 0.0000 9662690638.0470
> 9662690638.0470 1332198.0000 9662690638.0470 9662690638.0470
>
> MINIMIZER SLOWLY MOVING ATOMS WITH BAD CONTACTS DOWNHILL
> PRESSURE: 4 4.34298e+09 6.51726e+09 4.72519e+09 1.90218e+09
> 5.44552e+09 3.08174e+09 -2.88084e+07 -4.85553e+08 9.49706e+08
> GPRESSURE: 4 4.34298e+09 6.51726e+09 4.72519e+09 1.90218e+09
> 5.44552e+09 3.08174e+09 -2.88084e+07 -4.85553e+08 9.49706e+08
> ENERGY: 4 1583.3489 3638.1393 1260.5904
> 0.0000 -5655.4262 7235356386.7152 0.0000
> 0.0000 0.0000 7235357213.3676 0.0000
> 7235357213.3676 7235357213.3676 0.0000 3579402107.2727
> 3579402107.2727 1332198.0000 3579402107.2727 3579402107.2727
>
> to final
>
> WRITING EXTENDED SYSTEM TO RESTART FILE AT STEP 530
> WRITING COORDINATES TO DCD FILE AT STEP 530
> WRITING COORDINATES TO RESTART FILE AT STEP 530
> FINISHED WRITING RESTART COORDINATES
> WRITING VELOCITIES TO RESTART FILE AT STEP 530
> FINISHED WRITING RESTART VELOCITIES
> PRESSURE: 531 1.53654e+07 3.29332e+06 -372766 543512 1.72077e+07
> 1.39373e+06 1.05476e+06 1.19923e+06 1.78077e+07
> GPRESSURE: 531 1.53654e+07 3.29332e+06 -372766 543512 1.72077e+07
> 1.39373e+06 1.05476e+06 1.19923e+06 1.78077e+07
> ENERGY: 531 1600.1030 3648.7438 1294.4447
> 0.0000 -5651.9148 73889513.1000 0.0000
> 0.0000 0.0000 73890404.4767 0.0000
> 73890404.4767 73890404.4767 0.0000 16793614.9574
> 16793614.9574 1332198.0000 16793614.9574 16793614.9574
>
> WRITING COORDINATES TO DCD FILE AT STEP 531
> PRESSURE: 532 1.53654e+07 3.29332e+06 -372766 543512 1.72077e+07
> 1.39373e+06 1.05476e+06 1.19923e+06 1.78077e+07
> GPRESSURE: 532 1.53654e+07 3.29332e+06 -372766 543512 1.72077e+07
> 1.39373e+06 1.05476e+06 1.19923e+06 1.78077e+07
> ENERGY: 532 1600.1030 3648.7438 1294.4447
> 0.0000 -5651.9148 73889513.1000 0.0000
> 0.0000 0.0000 73890404.4767 0.0000
> 73890404.4767 73890404.4767 0.0000 16793614.9574
> 16793614.9574 1332198.0000 16793614.9574 16793614.9574
>
>
> Why the minimizer found it impossible to continue is what I am
> wondering about. I am considering:
>
> ---Before running conjugate gradient, run steepest descent (which is
> the rule in Amber). I rely on Peter's 2008 suggestion: "you can get
> something very similar by using velocityQuenching (turn on
> velocityQuenching and then use run X to run X steps). This method
> removes all velocity from all atoms at each step, which gives you a
> similar effect. See
> http://www.ks.uiuc.edu/Research/namd/2.6/ug/node29.html#8242 for more
> details."
>
> ---Increase Padding or Boundary, or both.
>
> ---Restrain also the water pertaining to the bilayer, in order to
> relax the external solvation water only (may be by setting 1.00 on col
> B for POPC water).
>
> ---Re-building protein+bilayer with a larger boundary (present model
> was built with 2.5A boundary between the protein and the bilayer, and
> could be minimized with namd under non-periodic conditions. This
> notwithstanding, may be that a larger boundary, 4.0A or so, is
> needed).
>
> That's all i can think about now.
>
> francesco
>
> On Tue, Dec 15, 2009 at 6:16 PM, Axel Kohlmeyer <akohlmey_at_gmail.com> wrote:
>
>> On Tue, 2009-12-15 at 17:01 +0100, Francesco Pietra wrote:
>>
>>> I forgot to ask: could you please suggest how roughly modify the
>>> parameters for cg solvation? I must have misinterpreted the analysis
>>> of inter-cell gap. In particular, the relationship between "Boundary"
>>> and "Box padding" is not clear to me.
>>>
>> please have a look at the online help of the solvate command. the
>> html file is a bit terse in that respect and should be updated.
>> also the URL pointing to the namd tutorial is off by one node...
>>
>> the boundary value is the distance between the solvent and solute.
>> the default value of 2.4 is fairly generous, but due to the increased
>> size of water (Martini rolls 4 waters into one site) and protein
>> side chain "atoms", stepping this up to, say, 4.0 might be a safe
>> choice. this can be easily rationalized from applying common sense:
>> just compare the values in an all-atom .par file to the coarse grain
>> .par file. the r2min/2 value in the CG .par file is 2.35 whereas the
>> corresponding AA values are between 1.3 and 1.8 with a few around 2.0.
>>
>> since your system was minimizing fine w/o periodicity, the default
>> might already be mostly ok for you. these values are empirical anyways.
>>
>> the padding value is how much solvent outside of min/max dimensions
>> of the solute you want to add. so, dimensions of solute _plus_
>> padding dimensions will be the new min/max of your system. AFAIK,
>> this does not include a "safety" (i.e. the equivalent of boundary for
>> inter cell distance), so i would just add that value or more to your
>> box dimensions.
>>
>> when visually checking for overlaps with PBC, you have to increase
>> the diameter of your vdw spheres (i just double them for our CMM
>> cg model).
>>
>> HTH,
>> axel.
>>
>>
>>> thanks
>>> francesco
>>>
>>>
>> --
>> Dr. Axel Kohlmeyer akohlmey_at_gmail.com
>> Institute for Computational Molecular Science
>> College of Science and Technology
>> Temple University, Philadelphia PA, USA.
>>
>>
>>
This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:53:37 CST