How to deal with Cu-as a ligand or as part of the receptor?

[Anonymous]

Nowadays,I try to model a Tyrosine,pdb name is 2y9w,there are two Cu exists int the protein. I checked the database and found that there’s no cu atom so there’s also no cu params file. So i trid to make a Cu params file manually, and add Cu atom in the database,but i don’t understand (NAME ATOM LJ_RADIUS LJ_WDEPTH LK_DGFREE LK_LAMBDA LK_VOLUME)in the file “rosetta-3.2/rosetta_database/chemical/atom_type_sets/fa_standard/atom_properties.txt”.what the data should i add for cu and what other files should i change. Besides,what should i deal with Cu,as a ligand or as a part of receptor?

[Anonymous]

I assume by the ligand/receptor distinction, you’re talking about using it in ligand docking. The answer to the question depends on what sort of question you’re asking. If you’re satisfied with the positioning of the Cu in the input PDB, you should treat the Cu atom as part of the receptor. The ligand docking code has no issues with having metal ions or even other small molecule co-factors as part of the receptor. On the other hand, if you want Rosetta to determine the location of the metal ion, you should treat it as a ligand. There’s two ways to go about this. One is to create a “hybrid” ligand which consists of the copper and your small molecule as a single molecule, connected by “dummy” bonds. I’d recommend using this approach if your copper is being chelated to a known location on the ligand. The other option is to use the multi residue ligand docking application/features, but I’m not sure if that is publicly released yet.

Regarding the copper parameters, the NAME and ATOM fields should be obvious. The LJ_RADIUS and LJ_WDEPTH fields are for the Lennard-Jones potential – radius and well depth. My understanding is that the Rosetta parameters were originally based off of the Charmm parameters, so you may be able to take the LJ parameters for copper from there. The LK_DGFREE, LK_LAMBDA, and LK_VOLUME terms are for the Lazaridis-Karplus implicit solvation model (See: http://www.ncbi.nlm.nih.gov/pubmed/10753811 ). I’m not sure if there’s been a parametrization of copper ions for the LK model, but my suggestion would be to copy the parameters from a “similar” ion.

Note that ideally you would want to validate whatever parameters you come up with, for example by coming up with a number of test cases that are similar to your problem of interest, but which have known answers. You see how well Rosetta does at recapitulating the known answers, and see if changing the copper parametrization would improve things. (Keeping in mind the issue of possibly overfitting to your benchmark set.) – But dealing with metal ions (even ones which have parameters in the database) isn’t something that Rosetta excels at currently, so you’re probably going to be mostly limited by other things, rather than the precise LJ & LK parameters for the ion.

[Anonymous]

Thank you for your answer!You said that there are two ways to deal Cu as a ligand.The first is to create a “hybrid” ligand.But if the Cu is isolated(means there is no small molecule),we should treat it as part of the receptor? If we treat it as part of receptor, we add the parameters data for Cu,then we can calculate the right score of the receptor consisting the Cu atoms,is that right?Another question:if we want treat a single Cu atom as a ligand,the params file and rotamer library for Cu should be created manually(because there is no method to deal with Cu in “molfile_to_params.py”),I create a params for Cu comparing to an exsited params file for Fe manually.Besides,I create a rotamer library for Cu which only consists of one rotamer(Because Cu is just a atom,),but when i use this as the input of ligand docking,the program will stop and show the message of”Missing atom positions in rotamer library!”,so i check the params file for cu and find there are two Virt atoms for Cu,and i add the position of two Virt atoms but there’s still the same wrong.Can you give some advice on how to treat a single icon as a ligand and how should i change the rotamer library for cu ? Thank you and hope for your reply!

[Anonymous]

Yes, if the copper is not directly in contact with the ligand, the “hybrid ligand” approach isn’t the one to take. If the Cu is isolated and you like the location of the copper ion in the receptor’s input PDB, treating it as part of the receptor is the way to go.

You are correct that you’ll need to make the params file for the copper manually, and copying an existing monoatomic metal ion is the way to do it. Note that if you only have one conformation (you don’t have any internal degrees of freedom you need to sample) you don’t actually need a rotamer library specified by the PDB_ROTAMERS line. The internal coordinates (ICOORD lines) in the params file should be sufficient for specifying the rotamer. (By the way, the virtual atoms are there due to a technical limitation of Rosetta. Rosetta needs to be able to establish a local coordinate frame for each residue, and it can’t do that with only one atom. It needs the (arbitrarily positioned) virtual atoms to lock down the rotational degrees of freedom for the residue.)

By the way, throughout this I am assuming that what you want to do is use the ligand docking code to dock an organic small molecule to a protein that also happens to contain a copper ion. If this isn’t what you’re looking to do, then you may need to approach things differently. Specifically, if you don’t have an organic small molecule, and are looking to dock a metal ion to a protein (i.e. determine the best xyz coordinates for the metal ion which results in the lowest energy of the complex), then Rosetta ligand docking might not the best way to go about doing it.

[Anonymous]

Thank you for your reply.Another question: if the Rosetta ligand docking might not be the best way to find the best xyz coordinates for a metal ion,what else method should I choose?

[Anonymous]

I don’t have any experience in looking for metal ion binding sites, but my recommendation is to look for a non-Rosetta program to do it, preferably one that specifically built for that purpose, or at the very least one which has been successfully used for that purpose in the past. I’m sorry that I don’t have any specific programs to recommend – it’s probably best to go through the literature looking for previously published reports of people doing what you want to do, and looking at what they use.

There are several problems with Rosetta and metal ions. First off, no one that I’m aware of has done extensive validation of metal ion binding in Rosetta. That’s not to say that metal ions aren’t used for some things: I believe ligand docking with metal ions in the binding site works about as well as other ligand docking programs, and people are working on enzyme & protein-protein association with designed metal binding sites. – But these either have the metal ion experimentally located, or have an “ideal” metal binding site with ideal geometry imposed on it externally. No one has really checked the parametrization of Rosetta metal ions to see if it’s able to pick out decent binding sites from non-binding sites (and definitely not for copper). They might, but you’d have to do extensive benchmarking to check first.

The other major problem is that Rosetta knows nothing about binding geometry. Almost all energy terms in Rosetta are residue-pairwise decomposable, so there’s nothing currently which would be able to specify that a particular metal ion likes an octahedral coordination geometry versus a tetrahedral versus a triganol bipyrimidal. It can’t even tell the difference between an ideal octahedral site with perfect 90 degree angles, and a distorted one where angles range all over the place from from 60-120 degrees. You might be able to add a new energy term, or put something together using virtual atoms on the metal ion, but that would likely take some development work and, again, extensive benchmarking.

So I’m not confident in Rosetta’s ability to use energy to discriminate between natural metal binding sites and spurious non-binding sites – it’s just not something that it’s currently set up and validated to do. You’re probably better off with a program that has literature precedent at being good at that sort of thing.

[Anonymous]

Thank you a lot for your information. I’ve tested that if I put the Cu in the pdb file,the score is different from the one without Cu. Is that the score not correct according to the message you told me:” Almost all energy terms in Rosetta are residue-pairwise decomposable, so there’s nothing currently which would be able to specify that a particular metal ion likes an octahedral coordination geometry versus a tetrahedral versus a triganol bipyrimidal.”? Or something else? Or because I’ve made the params file for Cu,so Cu is a residue logically,So Rosetta can calculate the score ?

[Anonymous]

Rosetta can calculate a score, it’s just not guaranteed to be a physically relevant score. In the broad brush strokes sense it is likely to give some indication of favorability, it’s just that if you have two poses, one with a copper ion score of -5 and one with a score of -7, I wouldn’t trust that -2 REU (Rosetta Energy Unit) difference to distinguish which one is a better binder.

This is the situation regarding coordination geometry. Since Rosetta doesn’t know anything about coordination geometry, all else being equal (interacting partners, distances, angles, clashes, etc.), it will score a tetrahedral coordinated metal ion the same as a square planar coordinated metal, even if that particular metal ion much prefers a square planar geometry. Likewise, it will score a distorted octahedron the same as a otherwise similar ideal octahedron, despite the fact that in nature there would be a large energy penalty for having a non-ideal geometry.

[Author]

Posts