[gmx-users] how to reimage PBC based on distance to a given selection without recentering or otherwise changing atomic coordinates

[Christopher Neale]

Dear users:

I have a system in which molecule A is in direct contact with molecule B. However, molecule B is imaged in a different periodic cell. What I would like to do is to get an image of both molecules in a periodic representation in which they actually are in contact (i.e., reimage molecule B such that it is closest to molecule A). However, I do not want to lose spatial information with e.g. a trjconv -center -pbc mol command. 

This is part of a complex automated build procedure and I can get into more details if that is useful, but the crux is that I am extracting a frame from a simulation, building more atoms onto molecule A, and setting up a new simulation. To build onto molecule A, I want to then do a vacuum EM before adding it back to the water box, for which I first enlarge the vacuum box, and changing box dimensions is messing with the periodicity and throwing molecule B away from molecule A in an unrealistic fashion (molecule B is a tightly bound ligand).

I could do what I want by breaking each molecule out into its own box, checking for contacts, reimaging, and then putting them back together. I presume (but have not checked) that I could also do this by making a new .itp file in which both molecule A and B are part of the same [ molecule ] definition and then running a zero-step mdrun. However, I am writing to see if anybody knows how reimage based on a selection (would be a single atom in molecule A near the contact between molecule A and B) more elegantly with processing tools available in gromacs.

Thank you for your help,
Chris.