- Posts: 15
MD run failing
- Fondyffendof
-
Topic Author
- Offline
- Fresh Boarder
Less
More
4 years 3 days ago #8503
by Fondyffendof
MD run failing was created by Fondyffendof
Hi everyone
I seem to be having some trouble with getting my system to run the actual md run.
I have energy minimised until the steps no longer change the energy of the system with some lincs warning:
It stopped because the algorithm tried to make a new step whose size was too
small, or there was no change in the energy since last step. Either way, we
regard the minimization as converged
but did not reach the requested Fmax < 1000.
Potential Energy = -1.9185824e+07
Maximum force = 3.9074165e+03 on atom 22514
Norm of force = 1.5074021e+01
The way i understand it is that lincs warnings are okay during the EM as the data is not used.
The equilibrium.mdp run leads to the same message as the EM: run is terminated due to too small step size with no change in energy...
My guess is that this means the system is equlibriated, do correct me if i am wrong.
running the md run then leads to lincs warning for 3 steps and terminates.
The system has been martinized from both gro and pdb both transformation leads to failure of the md run.
I am running in a cubic box with the .mdp files from the tutorials of DNA changing only the reaction field to PME and emtol =1000.
So my question is do anyone have any sugestions to what i could do to make it run?
I seem to be having some trouble with getting my system to run the actual md run.
I have energy minimised until the steps no longer change the energy of the system with some lincs warning:
It stopped because the algorithm tried to make a new step whose size was too
small, or there was no change in the energy since last step. Either way, we
regard the minimization as converged
but did not reach the requested Fmax < 1000.
Potential Energy = -1.9185824e+07
Maximum force = 3.9074165e+03 on atom 22514
Norm of force = 1.5074021e+01
The way i understand it is that lincs warnings are okay during the EM as the data is not used.
The equilibrium.mdp run leads to the same message as the EM: run is terminated due to too small step size with no change in energy...
My guess is that this means the system is equlibriated, do correct me if i am wrong.
running the md run then leads to lincs warning for 3 steps and terminates.
The system has been martinized from both gro and pdb both transformation leads to failure of the md run.
I am running in a cubic box with the .mdp files from the tutorials of DNA changing only the reaction field to PME and emtol =1000.
So my question is do anyone have any sugestions to what i could do to make it run?
Please Log in or Create an account to join the conversation.
- riccardo
-
- Offline
- Platinum Boarder
3 years 11 months ago #8510
by riccardo
Replied by riccardo on topic MD run failing
Sometimes the energy minimization convergence is not so telling.
Which time step are you using for the MD run? A useful thing to try always is to decrease the time step for the initial phase of the equilibration, and then slowly bringing it up. Note that there's no real recipe which works for all systems here, but you should rather go for some trial and error, e.g., start with 5 fs for a few ns, then 10 fs for a few more ns, then even larger time steps - even though at this point you should check which time step can be used with the Martini DNA model because that depends also on the type of elastic network you apply, if I'm not mistaken - please see pubs.acs.org/doi/abs/10.1021/acs.jctc.5b00286 .
Another possible cause of instabilities are thermostats and barostats - which ones are you using? The Berendsen ones are recommended for equilibration because they are more robust.
Which time step are you using for the MD run? A useful thing to try always is to decrease the time step for the initial phase of the equilibration, and then slowly bringing it up. Note that there's no real recipe which works for all systems here, but you should rather go for some trial and error, e.g., start with 5 fs for a few ns, then 10 fs for a few more ns, then even larger time steps - even though at this point you should check which time step can be used with the Martini DNA model because that depends also on the type of elastic network you apply, if I'm not mistaken - please see pubs.acs.org/doi/abs/10.1021/acs.jctc.5b00286 .
Another possible cause of instabilities are thermostats and barostats - which ones are you using? The Berendsen ones are recommended for equilibration because they are more robust.
Please Log in or Create an account to join the conversation.
- Fondyffendof
-
Topic Author
- Offline
- Fresh Boarder
Less
More
- Posts: 15
3 years 11 months ago #8512
by Fondyffendof
Replied by Fondyffendof on topic MD run failing
It is stable at 1 fs but blows up after a while with 2fs, which seems low for a cg simulation.
I am using v-rescale thermostat and Berendsen barostat with the stiff elastic network.
I am using v-rescale thermostat and Berendsen barostat with the stiff elastic network.
Please Log in or Create an account to join the conversation.
- Fondyffendof
-
Topic Author
- Offline
- Fresh Boarder
Less
More
- Posts: 15
3 years 11 months ago #8513
by Fondyffendof
Replied by Fondyffendof on topic MD run failing
And thanks for the information it helped
Please Log in or Create an account to join the conversation.
- riccardo
-
- Offline
- Platinum Boarder
3 years 11 months ago #8514
by riccardo
Replied by riccardo on topic MD run failing
Well, 2 fs is indeed too low for a CG simulation. v-rescale should also be quite robust so that's fine. You should be able to run Martini DNA with a timestep of 10 fs. How large is your system? What happens when it blows up (can you locate the cause by visualizing the trajectory and/or look at which atoms are involved in the constraints warnings) and how long is "after a while"?
Please Log in or Create an account to join the conversation.
- Fondyffendof
-
Topic Author
- Offline
- Fresh Boarder
Less
More
- Posts: 15
3 years 11 months ago #8515
by Fondyffendof
Replied by Fondyffendof on topic MD run failing
Well the system is quite large 1433750 beads .
The dna beads are part of the blow up after 700 ps.
In a triclinic 62 62 44 system.
The dna beads are part of the blow up after 700 ps.
In a triclinic 62 62 44 system.
Please Log in or Create an account to join the conversation.
- riccardo
-
- Offline
- Platinum Boarder
3 years 11 months ago #8516
by riccardo
Replied by riccardo on topic MD run failing
yeah, that's quite large. It might require some "massaging" to get that to run (a bit of what I was hinting at above, with slowly increasing time steps, etc.). I don't have direct experience with DNA (only with nucleobases) so I can't think of any specific issue commonly arising. I would look more closely into the beads blowing up and check whether there some some bad contacts/overlaps with other beads.
Also, pre-equilibrating parts of the system separately can also be a good idea: perhaps you have more than one DNA strand, plus lipids, proteins, don't know. You could equilibrate first the DNA strands on their own (e.g., in a box of water), and then put them back together in your bigger system. Depending on how your final system is composed this might require fancier procedures. Maybe in this chapter cgmartini.nl/images/practical-view-Martini-biomolsim19.pdf there are some smarter ideas.
Also, pre-equilibrating parts of the system separately can also be a good idea: perhaps you have more than one DNA strand, plus lipids, proteins, don't know. You could equilibrate first the DNA strands on their own (e.g., in a box of water), and then put them back together in your bigger system. Depending on how your final system is composed this might require fancier procedures. Maybe in this chapter cgmartini.nl/images/practical-view-Martini-biomolsim19.pdf there are some smarter ideas.
Please Log in or Create an account to join the conversation.
- Fondyffendof
-
Topic Author
- Offline
- Fresh Boarder
Less
More
- Posts: 15
3 years 11 months ago #8517
by Fondyffendof
Replied by Fondyffendof on topic MD run failing
Okay so equilibrate with 1 fs and try going up for the MD run makes sense.
I don't think that I can put the structure back together, if I equilibrate all the 84 strands separately, but I will remember this trick for any membrane interactions I might do thanks.
I don't think that I can put the structure back together, if I equilibrate all the 84 strands separately, but I will remember this trick for any membrane interactions I might do thanks.
Please Log in or Create an account to join the conversation.
- bart
-
- Offline
- Admin
Less
More
- Posts: 98
3 years 10 months ago #8564
by bart
Replied by bart on topic MD run failing
Did it work in the end? I was also wondering if turning PME off during equilibration might help. Also your books needs to be neutral when using PME, if this is not the case your system will most likely not be stable.
Please Log in or Create an account to join the conversation.
- Fondyffendof
-
Topic Author
- Offline
- Fresh Boarder
Less
More
- Posts: 15
3 years 10 months ago #8566
by Fondyffendof
Replied by Fondyffendof on topic MD run failing
Hi Bart
I found the problem with the way the dsDNA is read in the script.
The pdb containing my strands was written as individual single strands which the converter would connect in from the convertion.
Using the ssDNA model it worked, sorry for my mistake.
Using multiple equlibration i was able to simulate at higher timer steps thanks for the tip.
What do you mean by neutral books in the PME i might be confused with how the PME is working.
Best regards Chris
I found the problem with the way the dsDNA is read in the script.
The pdb containing my strands was written as individual single strands which the converter would connect in from the convertion.
Using the ssDNA model it worked, sorry for my mistake.
Using multiple equlibration i was able to simulate at higher timer steps thanks for the tip.
What do you mean by neutral books in the PME i might be confused with how the PME is working.
Best regards Chris
Please Log in or Create an account to join the conversation.
- vainikka
-
- Offline
- Admin
Less
More
- Posts: 91
3 years 10 months ago #8568
by vainikka
Replied by vainikka on topic MD run failing
Your system can't have a net charge when using PME, or you'll risk inducing artefacts (and potentially crashing your system). So for every charged particle you introduce, you will need to add an opposite charge. Usually done with counterions.
GROMACS will give you warning if you try and do this, so you should be aware.
GROMACS will give you warning if you try and do this, so you should be aware.
Please Log in or Create an account to join the conversation.
- Fondyffendof
-
Topic Author
- Offline
- Fresh Boarder
Less
More
- Posts: 15
3 years 10 months ago #8569
by Fondyffendof
Replied by Fondyffendof on topic MD run failing
That's what I figured, the net charge was balanced
Please Log in or Create an account to join the conversation.
- bart
-
- Offline
- Admin
Less
More
- Posts: 98
3 years 10 months ago - 3 years 10 months ago #8572
by bart
Replied by bart on topic MD run failing
My books should have been box. Having a neutral box is important. Sorry for the confusing typo ;) Nice that it all worked out!
Last edit: 3 years 10 months ago by bart.
Please Log in or Create an account to join the conversation.
Time to create page: 0.110 seconds