I'm using plumed 2.2 with Gromacs 5. I am currently struggling to cover a sufficient range of my three collective variables, two torsional (dihedral) and one distance.
My system essentially comprises of cylindrical proteins which must rotate around one another as a dimer pair. The rotation is determined via two dihedral angles built from the atoms of each protein e.g. p1_a,p1_b,p1_c,p2_d and p2_a,p2_b,p2_c,p1_d, notice the fourth part of the dihedral angle, it is the opposite protein. This gives a point of reference in an otherwise infinite environment. Also given that the rotation could happens as a monomer pair rather than a dimer, I am using a centre of mass distance between the two proteins. that makes up three collective variables.
I have tried regular meta-dynamics and well-tempered meta-dynamics. The regular meta-dynamics typically results in LINCS warnings and then finally a 1,4 error crashing gromacs. Whilst the well-tempered meta-dynamics doesn't cause the system to crash but it does take FOREVER and the bias factor usually reduces the gaussian potential hills too much.
I am struggling with setting my bias, height and sigma (width) of a system with no information regarding the energy landscape. From the various test simulations I am moving on the distance CV, starting from 1.4 and reaching 1.8 nm, but the two dihedral angles always fluctuate between -3 to +3. I have made a list of several parameters (SEE ATTACHED), with the highest metad.bias values being generated. As you can see from one of the metad.bias values, part of the free energy landscape within that CV range can reach up to 2248 kj/mol. Unfortunately, I don't seem to climb out of this part of the CV range.
Any thoughts or suggestions?
Thans
Anthony
Height 1 kj/Mol (even less), Sigma 0.35 rad for dihedrals and 0.05 nm for distances, is a pretty standard set up, with the biasfactor you can choose 15, or 20, or higher, but the simulation time you need to overcome the free energy barriers could be affected by orthogonal degrees of freedom which you are not biasing, and if you see hysteresis during the simulation it could mean that your dihedral CV is not a "good" CV, it could happen with pseudo - dihedrals. You can use replica-exchange methods to help your sampling, it will take care of the other barriers, using Bias exchange WT-MetaD or Parallel Tempering or Solute Tempering, Plumed and gromacs can do all these things, for defining "good" CVS, with physically relevant barriers, that's is more complicated, and very system dependent.
Height 1 kj/Mol (even less), Sigma 0.35 rad for dihedrals and 0.05 nm for distances, is a pretty standard set up, with the biasfactor you can choose 15, or 20, or higher, but the simulation time you need to overcome the free energy barriers could be affected by orthogonal degrees of freedom which you are not biasing, and if you see hysteresis during the simulation it could mean that your dihedral CV is not a "good" CV, it could happen with pseudo - dihedrals. You can use replica-exchange methods to help your sampling, it will take care of the other barriers, using Bias exchange WT-MetaD or Parallel Tempering or Solute Tempering, Plumed and gromacs can do all these things, for defining "good" CVS, with physically relevant barriers, that's is more complicated, and very system dependent.
Here is an example of pseudo-rigid orientation CVs for flexible molecules, with an example of a peptide, I'm not sure these CVs are implemented in PLUMED, but you can probably define them using MATHEVAL.
http://www.tandfonline.com/doi/full/10.1080/00268976.2013.813594
Hi Alejandro,
Could you please explain below statement more?
" if you see hysteresis during the simulation it could mean that your dihedral CV is not a "good" CV."
Thanks!
Hi, I am doing error analysis in metadynamics, but I don't get the metad.bias, I am only getting three column as in for the first COLVAR file. Any help.
Thanks
Hi Daniel,
Is your plumed input(plumed.dat) ok? Can you upload plumed.dat file? so that we can comment.
Thanks
anji
- Badges
- Science topic