Re: ABF simulations with singularities?

From: Junwoong Yoon (junwoony_at_andrew.cmu.edu)
Date: Mon Jul 09 2018 - 15:11:48 CDT

I'm running both metadynamics and umbrella samplings, but their output
ti.pmf values are so different (by order of one or two).
In metadynamics, the output is in range from 0~20.
In umbrella sampling, it's in range from 0~1000

Should I have to do any further post processing to fix it?
And what would happen if I increase the lower/upperWallConstant?

Thank you so much for your help
Jun

On Mon, Jul 9, 2018 at 2:00 PM, Giacomo Fiorin <giacomo.fiorin_at_gmail.com>
wrote:

> Hi Junwoong, as alternative to WHAM, since you are using distanceZ that
> supports ABF, you can also run a metadynamics or umbrella sampling run with
> the "writeTIPMF yes" option, which supported by other biases than ABF:
>
> See examples here:
> https://github.com/Colvars/examples/tree/master/tma-aco/smd-ti
> https://github.com/Colvars/examples/tree/master/tma-aco/mtd-ti
> https://github.com/Colvars/examples/tree/master/tma-aco/us-ti
>
> This is a new feature, and you will need a nightly build of NAMD.
>
> Giacomo
>
> On Mon, Jul 9, 2018 at 1:53 PM Junwoong Yoon <junwoony_at_andrew.cmu.edu>
> wrote:
>
>> Hi Jerome,
>>
>> I tried this out, but still the molecule seems stuck in the lower or
>> upper wall...
>> Should I try harder wall or more overlap between grid boundaries and the
>> harmonic walls?
>>
>> I also tried using Umbrella Sampling, but is there a way to reconstruct
>> PMF profile without using WHAM?
>>
>> On Fri, Jul 6, 2018 at 3:18 AM, Jérôme Hénin <jerome.henin_at_ibpc.fr>
>> wrote:
>>
>>> Hi,
>>>
>>> I wonder if the variable could be getting stuck in the "gap" between the
>>> boundaries of the force grid (beyond which there is no ABF bias) and the
>>> wall potential - structly speaking there is no gap, but here the wall
>>> potential is a little soft, so there is some space at the bottom of the
>>> harmonic well. That is especially likely if the free energy is decreasing
>>> towards the boundaries, which is an unusual situation because usually we
>>> sample until the edges of a basin, but is still possible.
>>>
>>> What you could try is have the grid boundaries overlap with the harmonic
>>> walls, and use a harder wall:
>>> colvar {
>>> name particleZ
>>> width 0.1
>>> lowerboundary 2
>>> lowerWall 3
>>> upperWall 30
>>> upperboundary 31
>>> lowerwallconstant 20
>>> upperwallconstant 20
>>>
>>> Then the rest would be the same. You can restart this simulation from a
>>> previous one to reuse the sampling you have. This will just extend the grid
>>> by 1 Angstrom on either side.
>>>
>>> For the purpose of understanding precisely what's happening in your
>>> existing simulations, you could try out this brand new visualization tool
>>> for VMD:
>>> https://github.com/Colvars/colvars/blob/master/vmd/scripts/cv_gui.tcl
>>> Load a trajectory in VMD, then type "play <path/to>/cv_gui.cl" in the
>>> console. Then you can load your Colvars configuration file, plot the
>>> trajectory of selected variables, and navigate it by clicking in the plot
>>> window.
>>>
>>> Best,
>>> Jerome
>>>
>>>
>>>
>>>
>>> On Fri, 6 Jul 2018 at 03:58, Junwoong Yoon <junwoony_at_andrew.cmu.edu>
>>> wrote:
>>>
>>>> Hi all,
>>>>
>>>> I'm doing a simple adsorption energy calculations of a surfactant
>>>> molecule onto a hard surface (graphene) in the presence of water molecules
>>>> (water box) using ABF (with distanceZ colvar). Le'ts say the original
>>>> z-direction coordinate of the surfactant as "z0", Then it only either goes
>>>> down toward lowerBoundary or goes up toward upperBoundary. Then the
>>>> simulation gets stuck near the boundary and does not explore the other way.
>>>> It seems like without solvation box, it worked fine (simulation could
>>>> explore the entire range in boundaries). However with the water solvation
>>>> box, this thing happens.
>>>>
>>>> Here's what I have been tried but could not resolve this issue:
>>>> 1) changing the grid limit (lower & upper boundaries)
>>>> 2) changing original height, z0
>>>> 3) increasing fullSamples value
>>>>
>>>> Please give me any recommendation if you ever seen this kind of issue
>>>> before or have any idea to resolve it.
>>>>
>>>> FYI, I have copied my colvars and namd configuration files here.
>>>>
>>>>
>>>>
>>>> ######### CONFIG FILE ##########
>>>>
>>>> cellBasisVector1 51.05 0.0 0.0
>>>> cellBasisVector2 0.0 51.579 0.0
>>>> cellBasisVector3 0.0 0.0 70
>>>> cellOrigin 0.0 0.0 0.0
>>>>
>>>> coordinates SDS_graphene_wb.pdb
>>>> #extendedSystem SDS_graphene_wb.xst
>>>> temperature 300
>>>> XSTfile ./output/cntabf
>>>> XSTfreq 1000
>>>>
>>>> set temp 300
>>>> set numstep 10000000
>>>>
>>>> switching on
>>>> switchdist 10
>>>> cutoff 12
>>>> pairlistdist 14.5
>>>> margin 10.0
>>>> stepspercycle 20
>>>>
>>>> #wrapWater on
>>>> wrapAll on
>>>>
>>>> #useFlexibleCell no
>>>> #useConstantArea no
>>>> #useGroupPressure no
>>>> langevinPiston on
>>>> langevinPistonTarget 1.01325
>>>> langevinPistonPeriod 200
>>>> langevinPistonDecay 100
>>>> langevinPistonTemp $temp
>>>>
>>>> langevin on
>>>> langevinTemp $temp
>>>> langevinDamping 0.5
>>>>
>>>> #This probably isn't necessary
>>>> reassignFreq 100000
>>>> reassignTemp $temp
>>>>
>>>> PME off
>>>> PMEGridSizeX 108
>>>> PMEGridSizeY 108
>>>> PMEGridSizeZ 80
>>>>
>>>> #Use structures for molecule and a small number of graphene parameters
>>>> structure SDS_graphene_wb.psf
>>>> paraTypeCharmm on
>>>> parameters par_all27_prot_lipid.inp
>>>> parameters par_cntgraph.inp
>>>> exclude scaled1-4
>>>> 1-4scaling 1.0
>>>>
>>>> timestep 1.0
>>>> #rigidBonds all
>>>>
>>>> fullElectFrequency 4
>>>>
>>>> numsteps $numstep
>>>> outputtiming 1000
>>>> outputenergies 1000
>>>> outputname ./output/cntabf
>>>> restartName ./restart/cntabf
>>>> restartFreq 1000
>>>>
>>>> fixedAtoms on
>>>> fixedAtomsFile SDS_graphene_wb.pdb
>>>>
>>>> DCDfile ./output/cnt_abf.dcd
>>>> DCDfreq 1000
>>>>
>>>> colvars on
>>>> colvarsConfig colvars_config_custom.tcl
>>>>
>>>> minimize 1000
>>>> reinitvels $temp
>>>> run $numstep
>>>>
>>>> ######### COLVARS FILE ################
>>>> colvarsTrajFrequency 1000
>>>> colvarsRestartFrequency 1000
>>>> smp off
>>>>
>>>> colvar {
>>>> name particleZ
>>>> width 0.1
>>>> lowerboundary 3
>>>> upperboundary 30
>>>> lowerwallconstant 5
>>>> upperwallconstant 5
>>>>
>>>> distanceZ {
>>>> #look this up in NAMD user guide
>>>> main {
>>>> atomsFile SDS_graphene_wb.pdb
>>>> atomsCol B
>>>> atomsColValue 1
>>>> }
>>>>
>>>> ref {
>>>> atomsFile SDS_graphene_wb.pdb
>>>> atomsCol B
>>>> atomsColValue 2
>>>> }
>>>> }
>>>> }
>>>>
>>>>
>>>> abf {
>>>> name densityABF
>>>> #timestepfactor 10
>>>> colvars particleZ
>>>> historyFreq 500
>>>> outputFreq 1000
>>>> fullSamples 500
>>>> }
>>>>
>>>>
>>>>
>>
>
> --
> Giacomo Fiorin
> Associate Professor of Research, Temple University, Philadelphia, PA
> Contractor, National Institutes of Health, Bethesda, MD
> http://goo.gl/Q3TBQU
> https://github.com/giacomofiorin
>

This archive was generated by hypermail 2.1.6 : Tue Dec 31 2019 - 23:20:04 CST