normal Cannot move past a very small time step

  • acnash
  • acnash's Avatar Topic Author
  • Offline
  • Junior Boarder
More
7 years 2 months ago #6851 by acnash
Hi all, this is a bit of a desperate plea in an attempt to get this working. I've been trying to bring my time step from the atomistic to the CG for well over a month now with no success.

I am using the identical settings from the latest .mdp:
www.cgmartini.nl/images/parameters/examp...tini_v2.x_new-rf.mdp

My constructions from an atomistic to CG has used the latest martinize and martini force field parameter set. I've looked through the whole .itp file of my protein and I haven't found any excessively strong bonds that should be turned into constraints.
I've started on an NVT with a time step of 0.00001 -> 0.0001 -> 0.002 using v-scale (see the latest .mdp file, above).

I then tried time steps of 0.005, 0.01, 0.02. None of them worked. So keeping with 0.002. I moved to an NPT (as in the .mdp file, but with a Berendsen p-coupling). This is worked indefinitely. I switched to Parnello pressure coupling with the small time step. Again, work fine. I've ran the system for up to 1 ns, with potential energy converging, with temperature and pressure oscillating quite happily around their target value. I tried moving to 0.025 but it through a lincs warning within the first two steps. So I moved to 0.003, and I see:

Energies (kJ/mol)
Bond G96Angle Proper Dih. Improper Dih. LJ (SR)
7.65298e+06 6.87737e+04 5.32698e+04 3.04081e+02 1.10664e+09
Coulomb (SR) Potential Kinetic En. Total Energy Temperature
-3.11753e+04 1.11438e+09 8.08514e+17 8.08514e+17 1.48146e+15
Pressure (bar) Constr. rmsd
2.14148e+15 1.70901e-04


Step 801 Warning: pressure scaling more than 1%, mu: 7.4074e+08 7.4074e+08 1.73075e+09

I am beginning to think that Martini is just not suitable for my protein. I understand that that is ridiculous, as the system does not matter. During my PhD I quite happily spent three years modelling CG transmembrane proteins in lipid bilayer, but I don't seem to be able to get human collagen from the atomistic to the CG. Perhaps it is my crystallographic unit dimensions:

4.31795 2.80443 337.85052 0.00000 0.00000 -0.78195 0.00000 0.14121 -0.32380

Please Log in or Create an account to join the conversation.

  • acnash
  • acnash's Avatar Topic Author
  • Offline
  • Junior Boarder
More
7 years 2 months ago #6852 by acnash
Replied by acnash on topic Cannot move past a very small time step
A further quick question, despite what is listed in the latest .mdp file (from Martini), which thermostat and pressure coupling should I be using? As an example, numerous sources over the past year have suggested that the pressure coupling should be Berendsen, but the latest .mdp file uses velocity rescaling. What should I stick to?

Please Log in or Create an account to join the conversation.

  • acnash
  • acnash's Avatar Topic Author
  • Offline
  • Junior Boarder
More
7 years 2 months ago #6853 by acnash
Replied by acnash on topic Cannot move past a very small time step
Sorry further question, I'm just adding here as I figure things out.

Judging by the atoms reported in the LINCS warnings and their respective position in the protein.itp file, I am having issues with constraints in the protein side chains e.g.,

The first LINCS error:
Step 3, time 0.06 (ps) LINCS WARNING
relative constraint deviation after LINCS:
rms 0.000028, max 0.001990 (between atoms 1620 and 1622)
bonds that rotated more than 30 degrees:
atom 1 atom 2 angle previous, current, constraint length
7771 7772 36.0 0.1100 0.1100 0.1100

In the .itp file:
1619 N0 811 PHE BB 1619 0.0000 ; F
1620 SC5 811 PHE SC1 1620 0.0000 ; F
1621 SC5 811 PHE SC2 1621 0.0000 ; F
1622 SC5 811 PHE SC3 1622 0.0000 ; F
.....
[ constraints ]
.....
1620 1621 1 0.27000 ; PHE
1620 1622 1 0.27000 ; PHE
1621 1622 1 0.27000 ; PHE

I suspect I have a topology problem, I just don't quite know how to fix it.

Please Log in or Create an account to join the conversation.

More
6 years 4 months ago #7420 by ananvodo
Replied by ananvodo on topic Cannot move past a very small time step
Hi Acnash.

I have a question. Have you changed your LINCS-warnangle to 90.
From my experience it doesn´t really creates any difference.

The problem you described happened to me with a one protein and will all others (fro the set I had) it did not.
From what I appreciate is that it depends on the compactness of the molecule itself and dssp file.

My advice is that you use ElNeDyn and for mdrun use the -rdd 1.4

Please Log in or Create an account to join the conversation.

Time to create page: 0.103 seconds