If you want to have faster speed without losing accuracy, you don't need to reduce the value of rgbmax
instead
1. Using dt=0.002 with any AMBER version. Or using dt=0.004 with H-mass repartining in AMBER14. (Check the manual)
2. Using gamma_ln = 1. This value control the 'viscosity' in GB simulation. Higher value decreases the sampling.
For Ubiqutin sysytem with dt=0.004, AMBER14 + 1 GPU, we were able to get >500ns/day (check the GPU bench mark in AMBER website or our paper).
Hai Nguyen
> On Oct 15, 2014, at 5:29 PM, Nihal Korkmaz <enihalkorkmaz.gmail.com> wrote:
>
> Thanks Hai,
>
> I am using the correct force field, gb and radii combination. Amber
> passed the tests as well. I just need to install the recent updates.
>
> I think the root of my problems is either my equilibration step or small
> rgbmax i have been using. I have been using 12 mainly to be able to run
> faster. I will check how much difference it makes and simulate with a
> larger rgbmax and equilibrate as described in the paper you mentioned.
>
> Thanks!
> Nihal
>
>> On 10/15/14, 12:00 AM, Hai Nguyen wrote:
>> Hi,
>>
>> it's likely there is something wrong with your protocol. I am not surprised
>> if protein unfold quickly in GB7 but we did have simulation for Ubiquitin
>> with 14SBonlysc + GB8 (PDB: 1UBQ) and it's very stable in our simulation at
>> 300K (>10 us). We have just published a paper showing using this force
>> field and solvent model with different topology, in both REMD and MD. You
>> might want to check out (see supporting information)
>> http://pubs.acs.org/doi/abs/10.1021/ja5032776
>>
>> You might want to double-check if you really used ff14SBonlysc or not. What
>> is radii for GB8 (should be mbondi3). Did the AMBER pass the test?
>>
>> Look at your your md input, I think you used very small value of
>> "rgbmax=12". You should use the default one.
>>
>> PS: did Ubiquitin unfold and turn alpha helix or just "unfold"?
>>
>> Hai
>>
>> On Tue, Oct 14, 2014 at 4:50 PM, Nihal Korkmaz <enihalkorkmaz.gmail.com>
>> wrote:
>>> Dear all,
>>>
>>> I have been running ff99SBnmr+gb7 and ff14SBonlysc+gb8 simulations. For
>>> only alpha helical systems I do not have any problems. However, with
>>> proteins that also contain B-sheets I observe high RMSD values and loss
>>> of secondary structure, mostly B-sheets. This is not only happening
>>> after simulating for a long time but I observe a steady increase in
>>> RMSD. For Ubiquitin monomer, I have 5 AA Calpha-RMSD at 70 ns, it
>>> reaches up to 10.5 AA and the Ub fold is lost.
>>>
>>> I am wondering if anyone experienced the same and has a suggestion to
>>> fix this.
>>>
>>> This is my input file:
>>> &cntrl
>>> imin=0, ntb=0, ntx=5, irest=1,
>>> ntpr=1000, ntwr=10000000, ntwx=1000, ntwe=5000,
>>> nstlim=10000000, dt=0.001,
>>>
>>> ntt=3, temp0=300.00, tempi=300.00,
>>> ig=-1, tautp=1, gamma_ln=20,
>>>
>>> ntc=2, tol=0.00001, ntf=2,
>>> dielc=1, cut=9999, rgbmax=12,
>>> igb=8, gbsa=0,
>>> saltcon=0.15, ioutfm=1, nscm=100,
>>> &end
>>>
>>> Thanks
>>>
>>> --
>>> E.N. Korkmaz
>>> UW-Madison, Biophysics Program
>>>
>>> _______________________________________________
>>> 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
>
> --
> E. Nihal Korkmaz
> UW-Madison, Biophysics Program
> Qiang Cui Lab
> Phone: 608-265-3644
> E-mail: korkmaz.wisc.edu
>
>
> _______________________________________________
> 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 Oct 15 2014 - 16:00:03 PDT
