I'm having some issues resolving artefacts generated during RMSF calculation of a protein from its crystallographic unit cell conditions. Using a combination of trjconv -pbc adjustments I'm able to make the molecule whole and almost all of it within the unit cell dimensions. Unfortunately, due to 1) the initial crystal structure position relative to the unit cell, and 2) that there isn't a single frame where the entire protein sits within the unit cell during a trajectory, I can't form a suitable trajectory for gmx rmsf without at least a tiny part of the protein structure crossing the unit cell.
The n-terminal residue of my protein always sits outside, and produces an RMSF (average over the residue) of 12.54 nm. Compared to the neighbouring residues with an RMSF of 0.2 to 0.7 nm, this issue is definitely down to the PBC.
Lots of gmx tools come with a -pbc flag. It is a shame RMSF does not. Any thought on how to get around this problem I'm having?
If the molecule has a hard time fitting in the unit cell all at once, aren't you concerned about minimum image violations? It seems you've set your unit cell to almost exactly fit the solute. This is generally really unwise; even a modest amount of rotation will have the molecule interacting with itself. Maybe that's by design?
In any case, recentering indeed should work, perhaps followed by -fit rot+trans to realign the molecule, assuming you have it oriented along the longest axis to begin with.
Ensure that the tpr file has the protein without any pbc issues. then try the following
gmx trjconv -f xtc -o whole.xtc -pbc whole
gmx trjconv -f whole.xtc -o cluster.xtc -pbc cluster
Hello Sarath,
As I said in my original post I have tried many different pbc configurations (I tried yours just to make sure), and whilst I am always able to keep my molecule whole, I always have part of the protein crossing the unit cell even whilst staying whole. See the image attached.
In a way, as I am looking at fluctuation, I almost want to centre to molecule with respects to the unit cell.
I've figured out a solution, and I'm putting it up here for anyone else who has a similar problem.
I only required one trjconv -pbc alteration applying mol and the -center option at the same time:
gmx_d trjconv -pbc mol -f wildtype_310.trr -s wildtype_310.tpr -o mol_center.gro -center -n index.ndx
I select the group I want to measure (see image above) for centering, and the output being the whole system. I then use the output file in both -s and -f of rmsf:
gmx_d rmsf -n index.ndx -b 1500 -f mol_center.gro -s mol_center.gro -od RMSD_310_A.xvg -res
It does throw a warning about pbc artefacts, however there are none visible and all the rmsf are identical to my earlier calculations apart from the residues that were crossing the unit cell which are now more in line with the other rmsf values.
If the molecule has a hard time fitting in the unit cell all at once, aren't you concerned about minimum image violations? It seems you've set your unit cell to almost exactly fit the solute. This is generally really unwise; even a modest amount of rotation will have the molecule interacting with itself. Maybe that's by design?
In any case, recentering indeed should work, perhaps followed by -fit rot+trans to realign the molecule, assuming you have it oriented along the longest axis to begin with.
Hi Justin, thanks for the reply. Recentering did the trick. With regards to the minimum image violation you are right, it is can be issue with this proteins (collagen) given how narrow they. My original image was a very poor one, there are actually 8 proteins (see attached) from the original crystal structure. They don't violate minimum image. Collagen exactly fits another collagen, with interstitial water in an area called the D-band. It is unlike globular proteins the unit cell fits around enough collagen molecules to prevent a single molecule from interacting with itself and thus you get the huge stacked 'rope' like structure. The delight of modelling fibril-like proteins!
- Badges
- Science topic