Hello,
I have a pdb with crystallographic waters which appear to be important for the protein structure, and perhaps function. I am generating mutations using pyrosetta and evaluting their relative scores, and exporting structures to be simulated with MD.
I would therefore prefer to leave the explicit waters and use the TP3.params from database/chemical/residue_type_sets/fa_standard/residue_types/water. The structure does not appear to have any issues or errors, but I have seen this note "Typically, you simply want to remove the water since PyRosetta does not use these for any application and they can cause problems." and want to be sure I don't run into such problems with this hybrid explicit/implicit solvent method (using ref2015 score function)
Thanks in advance for your feedback!
Terra
I believe the "typically" means that the problem is that you need a human to determine which waters you really want and keeping all crystalline waters is bad (worse a water box), hence why the generalisation against keeping waters.
In many protein you get a water mediated wannabe salt bridge (Lys···HOH···Glu) that improves the score greatly or waters that are part of catalysis, so those are the ones to keep. The whole water box is however a problem as it will likely blow up* and if you introduce a mutation that will affect the waters nearby.
*) One problem is that each HOH residue is separated by a jump, so you'd best set the ligand jumps to True (not a movemap default) or they will never move... but for a water box that means their escape.
There is a water framework in Rosetta called Spades. I played around with it in PyRosetta (cf. blog post) and albeit slow it very useful for watery protein–protein interactions. Majorly it can remove pointless waters. However, it is not meant to be used with a water box.
There is a cartesian MD within PyRosetta that works (e.g. this Q) with implicit solvent (and works with constrains --> steered MD), but is much slower than NAMD & co.