You are here

Re: Protein-RNA interface design

10 posts / 0 new
Last post
Re: Protein-RNA interface design


I am trying to run the protein-RNA interface design protocol present in rosetta_demos. I found that in the xml file the residues that are allowed for design are not present in the protein (1a9n_ABQ). Am I missing something here ??

I also want to know whether rosetta can score protein-RNA interactions well or not, and what extra terms are scored in calculating the binding energy of the designed interface in case of protein-RNA interactions?

Post Situation: 
Tue, 2015-04-07 04:44

First, keep in mind that the Rosetta demos aren't necessarily always kept up-to-date. For example, much of the muddling around with the database as indicated by the demo is no longer needed.

That's completely aside from your issue, though. The residues specified by the XML *are* in the PDB. It's just that the XML is being specified with pose numbering instead of PDB numbering. Pose numbering starts at one and increases by 1 for each residue, regardless of chain or insertion/deletion. It's not the same as PDB numbering, which is the number in the PDB. Normally you can tell the difference by the presence/absence of chain letter designations. If it has a chain letter with it, it's in PDB numbering. If it doesn't, it's pose numbering. (e.g. if you specified "residue 8", which residue 8 would it be? The one on chain A, the one on chain B, or the one on chain Q. If you're using PDB numbering, you need the chain letter to disambiguate.)

Regarding scoring, scoring of nucleic acid/protein interactions is one of the less developed parts of Rosetta. Take a look at the work coming out of the Rhiju Das lab from Stanford for the latest advancements. This is an older demo, so there may be recent improvements from newer papers you would want to take a look at.

Thu, 2015-04-16 09:58

Thank you rmoretti for your response. I have few unrelated questions. First, is any fast way of extracting pdbs from silent file as currently using extract.pbds.linugccrelease is taking too much time. I tried using the mpi version of that but I am getting some kind of error, while other mpi executables are running fine.

Second, I am running protein RNA docking run using ml script and I am getting Fnat score 0.0 for the all docked complexes. Could you tell me what could be the reason?

Third, I would like to know how to cluster the protein-RNA docked decoys using cluster application, as I getting errors something like CA atoms in pose not found ?

Lastly, which score should we plot for docking funnel in case of protein-RNA docking, is it total_score vs RMSD or I_sc (interface score) vs I_rms (Interface rmsd)? As far I understand in case of protein RNA docking its only the side chains at the interface which are optimized, so I_sc vs I_rms should be plotted. What do you think about this ??

Thu, 2015-04-23 23:37

extract_pdbs is probably going to be the fastest way to extract the PDBs. Note that if you only want to extract some of them, you can speed things up tremendously by supplying the tags for the ones you want with the "-in:file:tags" option.

One possibility for a zero Fnat is if there's some residue type set issue. Do you see the message "Fnat calc called with non-fullatom pose!!!" at all in the tracer messages? The other possibility is that there just aren't any contacts in your docked complex.

There's a very limited sort of autodetect in the cluster application for RNA - if the first residue in the structure is RNA, it switches to all-atom rmsd instead of CA rmsd. It doesn't look for RNA at any other positions, though. It also looks like there is an rna_cluster application, but I don't have any experience with it so I don't know how well it works with mixed RNA/protein structures. There's limited documentation at

Which score and which rmsd you use for score vs rmsd depends on what you want to test and see. To be honest, your best bet is probably to plot all four combinations and think about what the different plots are telling you.

Fri, 2015-04-24 12:56


I am using protein-RNA interface design tutorial for designing a protein-RNA interface. I would like to know how can I calculate the interface energy of each decoy in order to get the best decoy? Does interface analyzer works for protein-RNA complexes as well ??

Could someone recommend any other tool apart from Rosetta which can be used for protein-RNA interface design ??


Tue, 2015-05-26 07:11

I can't think of a reason a priori why the InterfaceAnalyzer wouldn't work on protein-RNA complexes.

The one caveat is that the default energy function isn't optimized for protein-RNA interactions, so you may wish to substitute a different scorefunction for doing the analysis.

Wed, 2015-05-27 09:17

Thank you for replying. Could you tell me (briefly) which scorefunction shall I use for protein-RNA and how it has to be replaced ?


Thu, 2015-05-28 00:47

Here's what Rhiju said when I asked him about protein-RNA interactions

          No tools are available or supported yet for protein-RNA interface design (or structure prediction)… our goal is to get benchmarks in place over the next year.

         For now, I’d recommend starting with: stepwise/stepwise_res_level_energy.wts which was put together as a starting point by a bunch of us at RosettaCon2014.

         If talaris2013 is taken as a starting point, probably should at least strengthen H-bonds by adding a flags:


(For the last, that means making a copy of the talaris2013.wts file and adding an additional line "NO_HB_ENV_DEP" to the file.)

Fri, 2015-05-29 12:40

Thank you for your response. But I am unable to locate both weight files in scoring/weights folder of rosetta3.4 database.

Mon, 2015-06-01 01:46

Yeah, Rosetta3.4 is about 3 years out of date at this point. Being relatively recent developments, the scorefunctions mentioned would only be available in relatively recent weekly releases.

Unless you're trying to recapitulate old work (in which case use the scorefunction they used), I'd recommend updating to a recent weekly release. (As mentioned, protein-RNA interaction is still something that's still rather primitive in current versions of Rosetta - the status in older versions of Rosetta would be that much more limited.)

Mon, 2015-06-01 15:27