[gmx-users] Crashes during protein-ligand simulation

[Justin A. Lemkul]

On 7/6/12 2:05 PM, James Starlight wrote:
> Justin,
>
>
> I've done all steps in accordance to your tutorial. I've already done
> the same systems with another ligands but had no problem.
>
> This time I've made topology of the ligand via ATB server. I've only
> noticed that some cgnr are too big in that topology . This is the
> example
>
> ADN     3
> [ atoms ]
> ;  nr  type  resnr  resid  atom  cgnr  charge    mass    total_charge
>      1    NT    1    ADN     N6    1   -0.844  14.0067
>      2     H    1    ADN    H11    1    0.422   1.0080
>      3     H    1    ADN    H12    1    0.422   1.0080      ;  0.000
>      4     C    1    ADN     C8    2    0.097  12.0110
>      5    HC    1    ADN    H01    2    0.177   1.0080
>      6    NR    1    ADN     N3    2   -0.642  14.0067
>      7     C    1    ADN     C4    2    0.175  12.0110
>      8     C    1    ADN     C5    2    0.092  12.0110
>      9    NR    1    ADN     N7    2   -0.556  14.0067
>     10     C    1    ADN     C6    2    0.657  12.0110      ;  0.000
>     11     C    1    ADN    C5'    3   -0.677  12.0110
>     12     C    1    ADN    C4'    3    0.834  12.0110
>     13    OE    1    ADN    O4'    3   -0.248  15.9994
>     14     C    1    ADN    C1'    3   -0.558  12.0110
>     15     C    1    ADN    C2'    3    0.603  12.0110
>     16     C    1    ADN    C3'    3   -0.212  12.0110
>     17    NR    1    ADN     N9    3    0.415  14.0067
>     18    OA    1    ADN    O2'    3   -0.606  15.9994
>     19     H    1    ADN    H08    3    0.482   1.0080
>     20    OA    1    ADN    O3'    3   -0.606  15.9994
>     21     H    1    ADN    H06    3    0.482   1.0080
>     22    OA    1    ADN    O5'    3   -0.246  15.9994
>     23     H    1    ADN    H03    3    0.337   1.0080      ; -0.000
>     24     C    1    ADN     C2    4    0.502  12.0110
>     25    HC    1    ADN    H10    4    0.106   1.0080
>     26    NR    1    ADN     N1    4   -0.608  14.0067      ;  0.000
> ; total charge of the molecule:  -0.000
>

Large charge groups could account for errors in neighbor searching, leading to 
clashes that cause the simulation to collapse.

>
> 2) To the binding pocket I've inserted this ligand manually by means
> of superimposition with the reference x-ray structure wich include the
> same protein in the same conformation with the same ligand. I've done
> some systems already and that aproach was good :)
>

OK, just be ready for reviewers to ask why you didn't do docking ;)

> 3) It's strange that the simulation crashes without any reasons ( the
> system is very stable during calculated 10-15ns trajectory)
>

There's always a reason, you just haven't found it yet.  The charge group size 
could indeed be the problem; neighbor searching can fail at any time when some 
atoms run into one another.

> Also I suppose that such problems could be with the COM groups
>
> this is the example from my mdp
>
> comm-grps       = SOL_NA_CL XW Protein_CCl4_ADN
>
> here XW is the water wich were coppied from X-ray structure .
> Also in that system Ccl4 is the membrane mimicking layer so I've
> merged it with protein and ligand in the same group.
>

I see no reason to add such complexity to the system.  Breaking the crystal 
waters into their own COM removal group does not make sense to me.  Physically, 
they are basically part of the protein.

> On the current stage I've tried to make changes in the mdp on
>
> comm-grps       = System
>
> to check if the problem was with that COM motion
>

And what was the outcome?  I see no reason that two COM motion removal groups 
wouldn't be appropriate (as layers can slide with respect to one another, like a 
membrane) but three groups does not sound appropriate.

-Justin

-- 
========================================

Justin A. Lemkul, Ph.D.
Research Scientist
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin

========================================