You are here

De novo backbone trace from fragments

3 posts / 0 new
Last post
De novo backbone trace from fragments

Dear Rosetta experts,

I have been using rosetta mainly for refinement and iterative local rebuilding into cryo-em densities at reolustions of ~3-3.5 Angstrom. I am wondering how (or if it is possible) to combine and complete a set of backbone fragments predicted by external software into a single backbone trace using rosetta. To make matters somewhat more complicated I would need to do this without any sequence information using only polyA fragments.

I can generate multiple backbone fragments from density alone using buccaneer. The fragments have a length of 5-25aas and a high local accuracy but poor overall correlation with the native structure. Superimposistion of all these fragments reveal that a "consensus" of structure could be quite accurate even without sequence information. I have looked at iterative RosettaCM modelling as described in the electron density refinement tutorial. This however seems to rely quire heavily on sequence alignment for merging and completeing models. I am not sure how I would, or if I could, implement this for a set of predicted polyA stretches. 

The end goal would be to try to automate the intial chain tracing for large, multichain cryo-em structures where protein identity is determined from density alone. 

Any help or insight would be appreciated.

Best regards,

Victor Tobiasson

Post Situation: 
Tue, 2020-12-22 06:34
Victor Tobiasson

Hello Victor,


As I understand it, you essentially have a CA trace w/ your density and are hoping that your CA trace could be used in order to assign residues in your map.


If that is correct I would suggest following the de-novo modeling tutorial but using your CA trace and adding the flag "-ca_positions" as described at and supplying your polyA map.  I have seen this help in the past so I know it has the potential to work!


in terms of automating this, or doing this where there a lot of chains etc.  That is a very difficult problem and is something we are actively researching! If you get stuck and are completely out of ideas you should contact Frank DiMaio at UW we may have some unpublished applications that could help.

Tue, 2020-12-22 08:58

Hello, thank you for your prompt reply!


"As I understand it, you essentially have a CA trace w/ your density and are hoping that your CA trace could be used in order to assign residues in your map."

Not quite. I can generate a series of partial C-alpha traces all which fits parts of the density. None of them are perfect but on average they describe the local structure very well. I am wondering if there is a way of using these individual partial traces to creating a sort of consensus model, similar as to using several different homology models to create a consensus structure. Problem here being of course that there is no sequence alignment to go with. 

Thank you for providing the tutorial link. I think section 4C is similar to what I have in mind. The assembly of a series of probable substructured into a “maximally consistent” structure. Problems is as mentioned, it needs to be done with polyA.  


I attached an image to perhaps explain better. 


File attachments: 
Tue, 2020-12-22 09:18
Victor Tobiasson