[gmx-users] Advice for simulating small DNA

[Mark Abraham]

Joshua Ballanco wrote:
> On Feb 2, 2009, at 12:26 AM, Mark Abraham wrote:
> 
>> No, I've no idea since I don't simulate DNA.
> 
> In that case, thank you the help that much more!
> 
>>>>>>> So I'm now attempting to add restraints for the base-pair 
>>>>>>> H-bonds, but I'm having trouble. It seems like no matter what I 
>>>>>>> try, my system reliably explodes within the first 1 ns. My 
>>>>>>> constraints look like this:
>>>>>>> [ distance_restraints ]
>>>>>>> ; ai  aj  type  index type’ low up1 up2 fac
>>>>>>> 18  136 1     0     2     0.0 2.0 2.1 1.0
>>>>>>> 14  134 1     0     2     0.0 2.0 2.1 1.0
>>>>>>> 43  114 1     0     2     0.0 2.0 2.1 1.0
>>>>>>> 39  112 1     0     2     0.0 2.0 2.1 1.0
>>>>>>> 68   92 1     0     2     0.0 2.0 2.1 1.0
>>>>>>> 64   90 1     0     2     0.0 2.0 2.1 1.0
>>>>>>> I've tried pre-equilibrating for up to 100 ps, but even that 
>>>>>>> doesn't prevent the system from eventually exploding.
>>>>>>
>>>>>> Your .mdp settings for distance restraints may also be relevant 
>>>>>> here - not least in setting the existence and magnitude of these 
>>>>>> restraints.
>>>>> As I understand, the only relevant lines are:
>>>>> constraints         =  all-bonds
>>>>> integrator          =  md
>>>>> disre               =  simple
>>>>
>>>> disre-fc and others are also relevant. See manual chapter 7.
>>> Thanks for the pointer. I had overlooked most of the options there, 
>>> since I'm not actually doing anything related to NMR. (That'll teach 
>>> me to read more carefully!) Unfortunately, playing around with this, 
>>> disre-tau, disre-weighting, and the weighting factors for each bond 
>>> have not, so far, avoided the explosion.
>>
>> OK, that's no longer surprising - distance restraints will not 
>> usefully fix a broken model physics.
> 
> Well, yes, but I also wouldn't expect them to break the broken physics 
> further... I realize the system I was using originally was rather 
> unphysical, but the DNA helix at least was at least *mostly* holding 
> together. When I add the distance restraints, even with very large 
> multipliers, the seem to serve only to tear apart the helix. Odd...

Shrug. Resonance can do weird things - just ask the Tacoma Narrows bridge!

Mark