# Re: [AMBER] live dCAT simulation

From: Bill Ross <ross.cgl.ucsf.edu>
Date: Wed, 11 May 2016 21:39:27 -0700

E.g. a negative value for nscm could cause recentering on the origin as
well as removal of motion.

Bill

On 5/11/16 9:25 PM, Bill Ross wrote:
> I was looking at the wrong atom.
>
> Turning off SHAKE solved the immediate problem (dt=0.001). nscm and
> tautp also come in handy when one is throwing random forces into the
> equation.
>
> Is there any way to recenter the molecule on the origin? It could save
> removing center of mass motion at every step for my infinite simulation,
> which I expect will drift anyway over the centuries.
>
> Bill
>
>
> On 5/11/16 3:05 AM, Bill Ross wrote:
>> I tried adding these lines to multiply an atom's velocity in mdwrit.F90
>>
>> ! ATOM 16 C4 C5 1 6.021 2.830 -3.917 1.00
>> 0.00 C
>>
>> ----
>>
>> if (mod(nstep,5000)==0) then
>> atnum = 1 + (16-1) * 3
>> write(6, *) ' multiplying velocities at ', nstep
>> v(atnum) = 25.0 * v(atnum)
>> v(atnum+1) = 25.0 * v(atnum+1)
>> v(atnum+2) = 25.0 * v(atnum+2)
>> end if
>>
>> if (ntwr >= 0) return
>>
>> ----
>>
>> and crashed things, so I know I'm having an effect,
>>
>> vlimit exceeded for step 5041; vmax = 21.8359
>>
>> However, with ntwx=1, I can't see anything obvious happening as I step
>> through in VMD.
>>
>> Maybe this is a limit of numerical stability, that one can't inject a
>> visible amount of velocity?
>>
>> Thanks,
>>
>> Bill
>>
>>
>> On 5/10/16 10:39 PM, Bill Ross wrote:
>>> I'm not restarting, I'm just using the program's pause to write restrt
>>> coords (once per second) to also read the forces from the 'tap files'
>>> and apply them. I.e. I'll add some code to sander for this. If anyone
>>> has any other design considerations that would make it more generally
>>> useful, I'm all ears.
>>>
>>>
>>> On 5/10/16 8:55 PM, Charles Lin wrote:
>>>> If you're restarting the simulation each time the restart coordinate is getting printed, and you want to minimize adding functions to sander it would probably be faster to just use ntx=5 and change the velocity value in the restart file rather than modifying the forces.
>>>> ________________________________________
>>>> From: Bill Ross [ross.cgl.ucsf.edu]
>>>> Sent: Tuesday, May 10, 2016 8:09 PM
>>>> To: AMBER Mailing List
>>>> Subject: Re: [AMBER] live dCAT simulation
>>>>
>>>> I'm actually using clicks from my AI app running separately. Given the
>>>> possible issues in modifying NMR coordinates, maybe for each click I'll
>>>> just choose an atom and a force vector within the AI app, drop it into a
>>>> file, and have sander pick up the click files when it writes restrt, and
>>>> modify whatever it needs to modify. That last step is what I need to
>>>> figure out next.
>>>>
>>>>
>>>> On 5/10/16 7:59 PM, Charles Lin wrote:
>>>>> I would probably just augment the force array by # of clicks by modulating the x,y,z pixel coordinates of the mouse click with number of atoms and make each click augment force on that atom by 0.0001 or something small enough that a large amount of clicks would influence movement, but not enough that it would break the simulation. This would probably be the easiest hack because you can use atm_id and just affect the force array.
>>>>>
>>>>> I'm not entirely sure if adjusting the weights on the nmr code would achieve what you want. Perhaps changing the actual reaction coordinate values for the nmr code and not the energy barriers might work.
>>>>> ________________________________________
>>>>> From: Bill Ross [ross.cgl.ucsf.edu]
>>>>> Sent: Tuesday, May 10, 2016 7:42 PM
>>>>> To: AMBER Mailing List
>>>>> Subject: Re: [AMBER] live dCAT simulation
>>>>>
>>>>> I added updates of temp and energy and time, which latter I hope to see
>>>>> eventually overflow:
>>>>>
>>>>> http://phobrain.com/pr/home/mol.html
>>>>>
>>>>> Reload page if no molecule shows.
>>>>>
>>>>> Still thinking about how to inject live forces.. are people busy working
>>>>> on grants?
>>>>>
>>>>> Bill
>>>>>
>>>>>
>>>>> On 5/5/16 4:52 PM, Bill Ross wrote:
>>>>>> I've put up a web page showing a live simulation of dCAT in Generalized
>>>>>> Born solvent, using sander and cpptraj, with Alexander Rose's NGL
>>>>>> package on the browser side. (dDOG was not amused :-)
>>>>>>
>>>>>> http://phobrain.com/pr/home/mol.html
>>>>>>
>>>>>> It renders at one frame per second, rms fitted by cpptraj to the middle
>>>>>> base. There is an informal info page telling how to set up the Amber
>>>>>> side of things, which might be useful for teaching, and I give the code
>>>>>> for the web server side for people who want to broadcast their own live
>>>>>> dynamics. Suggestions for improvement are welcome, within my framework
>>>>>> of hillbilly science. A missing bond on the molecule display is a known
>>>>>> problem.
>>>>>>
>>>>>> The simulation's main purpose is to provide correlated noise to my slide
>>>>>> show image selection algorithms for AI purposes, which it does through
>>>>>> cpptraj base angle measurements. Now I would like to translate user
>>>>>> clicks in this entirely different domain into taps on the simulated
>>>>>> molecule (or perhaps decaying forces). So I'll need to check a directory
>>>>>> for tap files at some intervals (or possibly a socket). Probably when
>>>>>> restrt is written.
>>>>>>
>>>>>> Can any developers chime in on what I'd do next? An easy thing might be
>>>>>> to rescale some forces, chosen anonymously based on indexing, but it
>>>>>> would be nice to be able to formulate residue- and atom-level
>>>>>> descriptions to direct things more specifically. NMR facilities come to
>>>>>> mind, but I'm not sure how computationally expensive it would be to
>>>>>> invoke them, nor have I had anything do with NMR code. Maybe pre-define
>>>>>> a bunch of restraints, then dynamically tweak their weights, though it's
>>>>>> be nice in general to enable a reach-out-and-touch-anything interface. I
>>>>>> don't know if any folk want to take a step in the direction of having
>>>>>> live interaction with sander, but if so maybe it's worth thinking
>>>>>> through the interface and the design together, and incorporating the
>>>>>> code I come up with. (The feature is already in NAMD and LAMMPS.. maybe
>>>>>> I'll even wind up back on the author list :-)
>>>>>>
>>>>>> Thanks,
>>>>>>
>>>>>> Bill
>>>>>>
>>>>>> -- How it's done
>>>>>>
>>>>>> Sander writes to restrt every second, and a background script runs
>>>>>> cpptraj to RMS fit on 3 A base atoms every second. The javascript in the
>>>>>> browser requests coordinates every second, and a Servlet on the web
>>>>>> server picks up the current coordinate set and sends it to the
>>>>>> javascript, which gives it to NGL to render (NGL loads a pdb at first,
>>>>>> to figure out how it will render).
>>>>>>
>>>>>>
>>>>>>
>>>>>> _______________________________________________
>>>>>> AMBER mailing list
>>>>>> AMBER.ambermd.org
>>>>>> http://lists.ambermd.org/mailman/listinfo/amber
>>>>> _______________________________________________
>>>>> AMBER mailing list
>>>>> AMBER.ambermd.org
>>>>> http://lists.ambermd.org/mailman/listinfo/amber
>>>>>
>>>>> _______________________________________________
>>>>> AMBER mailing list
>>>>> AMBER.ambermd.org
>>>>> http://lists.ambermd.org/mailman/listinfo/amber
>>>> _______________________________________________
>>>> AMBER mailing list
>>>> AMBER.ambermd.org
>>>> http://lists.ambermd.org/mailman/listinfo/amber
>>>>
>>>> _______________________________________________
>>>> AMBER mailing list
>>>> AMBER.ambermd.org
>>>> http://lists.ambermd.org/mailman/listinfo/amber
>>> _______________________________________________
>>> AMBER mailing list
>>> AMBER.ambermd.org
>>> http://lists.ambermd.org/mailman/listinfo/amber
>> _______________________________________________
>> AMBER mailing list
>> AMBER.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber

_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed May 11 2016 - 22:00:03 PDT
Custom Search