From: Bennion, Brian (Bennion1_at_llnl.gov)
Date: Mon Feb 28 2011 - 16:06:28 CST

Hello Peter,

there is a general exception to the subselection overcounting issue, one of which is when waters in contact with a protein or a portion of it.

water within 5 of resid 3 and protein

will only display water and resid 3.

or the reverse,

protein within 2 of water

will only show the protein atoms and not the water.

So I have things figured out now and running. I appreciate the help.
I will also send you a slightly revised tutorial page for the plugin.

brian

________________________________________
From: Peter Freddolino [petefred_at_ks.uiuc.edu]
Sent: Friday, February 25, 2011 6:09 PM
To: Bennion, Brian
Cc: vmd-l_at_ks.uiuc.edu
Subject: Re: vmd-l: vmd1.9beta1 namdEnergyPlugin

Hi Brian,

In general, if you are purely using the atomselect language, "within 5
of (subselection)" will *always* include subselection. Any plugin where
this does not occur is taking care of removing the overlap on its own.

I cannot speak for resname SSON, or whether it is classified as protein,
for your particular case (since I don't have access to it), but when I
tried reproducing the problem that you described, the only way that I
could get it to occur was when the selections were nondisjoint due to a
within selection in sel2 that also included sel1. The best way to see
whether this is happening in your case is to make atom selections
corresponding to the text that you fed to namdenergy and see what they
contain. If this is not the source of your problem please let me know.

Best,
Peter

On 02/25/2011 11:45 AM, Bennion, Brian wrote:
> Thank you Peter for the reply.
> Yes I was confused on the skip switch and it does in fact work as stated.
>
> The atomselection is still a bit puzzling, just want to confirm what you said. There are some instances in vmd when a group of atoms is selected based on a distance to some other reference object(s), the reference object is not included. From your description below, it sounds like the selections I make in the plugin are including the reference object(s) in its selection because the resname SSON is also part of the protein. Is that correct?
>
> Thank you for the fixes.
>
> Brian
>
>
> -----Original Message-----
> From: Peter Freddolino [mailto:petefred_at_ks.uiuc.edu]
> Sent: Thursday, February 24, 2011 7:04 PM
> To: Bennion, Brian
> Cc: vmd-l_at_ks.uiuc.edu
> Subject: Re: vmd-l: vmd1.9beta1 namdEnergyPlugin
>
> Hi Brian,
>
> Regarding switch/skip, your understanding of stride is correct in terms
> of how it is documented and intended, but the timestep and stride
> functionalities appear to be broken for some reason. I have just patched
> them, so they should now work in the cvs version. However, in my tests
> skip **is** working as documented -- skip n causes calculation only for
> every nth frame, it does not simply skip n frames at the beginning.
> Could you confirm that you see unexpected behavior here?
>
> Regarding the selections, I think the problem here is that the atoms in
> selection 1 are also part of selection 2. Because pairinteractions in
> namd require interacting atoms to be flagged with a 1 or a 2 in the
> occupancy or beta field to correspond to group 1 or group 2, you end up
> with zero atoms in group 1 because those atoms are also in group 2. This
> also explains why things work ok when you switch the order of the
> selections. In the fix I just committed there is now a check for this
> situation, which should give an informative error when it occurs.
>
> Please let me know if this fixes things.
> Best,
> Peter
>
> On 02/24/2011 01:42 PM, Bennion, Brian wrote:
>> Hello Peter,
>> Thank you for your reply. Here are more details of my calculations.
>>
>> First, I load in 1 pdb file 1 dcd file with stride 10. So I have 201 total structures. For testing purposed I just wanted to take a small sample of that 200, say every 10 structures for a total of 20. It would appear that stride in the gui would do this for me. However, if I understand what you are saying, is that this switch is only to correlate the labeled timestep with whatever the original timestep was for the dcd file I read it. That is, since I already did used a "real" stride of 10 when reading the dcd file into vmd then I should set stride in the gui for 10 to keep timesteps on track. Is that the correct understanding? The namdgui stride switch does not do a further downselect of frames? Skip just blows through X frames and then starts up at the regular granularity of 1.
>>
>> Unfortunately I cannot send a structure at the moment to help address the second selection problems.
>> But as I stated in the first email. The only selections that I have been able to get to work are single residue pairs. For instance if the first selection field contains: resid 203 and protein and the second selection field contains: resid 124 and protein. I get the appropriate plots and data files showing the forces and energies. However, if the second selection contains: protein and same residue as within 6 of resid 203 and protein then the calculations "proceeds" but at the end, the plot and data file contain only zeroes.
>>
>> Can you send me your example and I can test it?
>>
>> Brian
>>
>>
>> -----Original Message-----
>> From: Peter Freddolino [mailto:petefred_at_ks.uiuc.edu]
>> Sent: Wednesday, February 23, 2011 5:30 PM
>> To: Bennion, Brian
>> Cc: vmd-l_at_ks.uiuc.edu
>> Subject: Re: vmd-l: vmd1.9beta1
>>
>> Hi Brian,
>> I think there may be some confusion about the usage of stride, skip, and
>> step length. Stride and step length are purely for bookkeeping purposes,
>> and help make sure that the times listed in the output match up with
>> your real timestep. What you're after is probably the skip argument.
>>
>> Regarding selections, that sort of thing should work just fine, and
>> indeed, does when I try it on typical test systems. Could you give an
>> exact example of input structure and arguments where it fails?
>>
>> Thanks,
>> Peter
>>
>> On 02/23/2011 03:21 PM, Bennion, Brian wrote:
>>> It would appear that there is a small problem with the namdenergy plugin. The gui allows for a stride entry to be made, however, it appears to be ignored when you receive the eventual energy/force plot. For example, I have 201 frames loaded in vmd and with a stride setting of 100 I still have 200 data points in the plot. This behavior is the same in vmd1.8.7.
>>>
>>> In addition, it would also seem that a "protein and same residue as within 6 of resname blah" in the second selection is not supported. Only values of zero are reported in the final plot and data file. Is the plugin supposed to be able to accept this type of atom selection query?
>>>
>>> Thanks
>>> Brian Bennion
>>>
>>>
>>>
>>>
>>>
>>>
>>> Biosciences and Biotechnology Division
>>> Lawrence Livermore National Laboratory
>>> 7000 East Ave Livermore California 94550
>>> phone:925-422-5722
>>> fax:925-424-4334
>>>
>>>