Altering Score/EMapVector

Why do you want to assign values to a list of scores, or for what purpose do the scores need to be modified? Rosetta tells you what the scores of the pose are; you can’t tell it what the scores are because it is the thing that does the scoring.

Another thing to point out is that the Energies object in a pose is really just a storage container. Anything that causes that residue to move (or even residues which are in energetic contact with that residue) will invalidate that cached value, and cause it to be recalculated, ignoring and discarding whatever value is there currently. Packing and minimization have their own energy caching mechanisms, which I believe are distinct from the Energies object, so they won’t be affected by changing that value. So even if you could change the value, it would quickly be invalidated, and really wouldn’t affect most of Rosetta’s internal mechanisms.

If you want to adjust the results of an energy calculation, the best way to go about doing it is to define your own custom energy term. There’s a way to do this through the Python interface, at least with PyRosetta 2.011 (“Examples can be found in test/*Subclassing*”)

What you want to do makes sense now.

Monte Carlo will re-score the Pose if it detects the scores are out-of-date, so you cannot do it in the way you envision. You might get it to work if you altered a lot of the Energies and MC object, but it’s dangerous.

There are two options I can think of: First, do your MC control without the Monte Carlo object. Basic MC control is not very hard to code yourself into your cycles-loop. This will let you alter scores as you please.

Second, you could detect the whatever-you’re-detecting and add constraints to alter the scorefunction. Constraints are evaluated on a per-residue basis as you seem to need. Don’t forget to dump the previous cycle’s constraints.

That’s the benefit of writing a custom energy term – you can check the residues you’re scoring against whatever list you’re using, and only put in non-zero energies for those residues you want to upweight. (i.e. if the residue or residue pair isn’t on your list, just give a value of zero for that residue/residue pair, and let the regular energy term handle the non-unweighted value). You can also incorporate the detection logic for which residues are on the list in the scoreterm object itself, automating the detection on repacking.

I’m not quite sure what you’re trying to achieve, but it sounds similar to protocols.enzdes.ProteinLigandInterfaceUpweighter, which specifically upweights the (packer) interactions between a ligand and surrounding residues. It uses a core.pack.task.IGEdgeReweighter and core.pack.task.IGEdgeReweightContainer (IG = interaction graph) to do the job. Specifically, it looks like it uses a protocols.toolbox.IGLigandDesignEdgeUpweighter, but there is also a protocols.toolbox.ResidueGroupIGEdgeUpweighter for increasing the weights between groups of residues. Best I can tell, though, those are for increasing the weights on all the scoreterms, and only work with the packer. I’m not sure if there is a similar general solution for up-weighting specific residue interactions in all instances

I can’t tell if it’s in your version of PyRosetta, but in the C++ version, the protocols::moves::MonteCarlo object has two boltzmann() member functions, one which takes a pose, and another which just takes a score delta. You could possibly use the latter instead of the former, and then keep track of the accepted Poses separately. That way you can do whatever score manipulation you need prior to passing it to the MonteCarlo object. I don’t know how that function version is exposed under the PyRosetta interface, though.