You are here

Can the Rosetta Antibody Design (RAbD) protocol incorporate glycans?

6 posts / 0 new
Last post
Can the Rosetta Antibody Design (RAbD) protocol incorporate glycans?


I would like to use RAbD to optimize an antibody that binds to an epitope that includes a glycan. Does RAbD support including glycans as part of the eptiope? 


Post Situation: 
Tue, 2020-05-05 20:30

As the developer of RAbD and a major collaborator on RosettaCarbohydrates - it should, but has not been tested. 

The glycans cannot be part of the CDR loops, but should repack.  They will not minimize during RAbD, however, so it's probably best to start with an ensemble of glycan structures.  You could also model the glycan afterwards using the GlycanTreeModeler. If you want to chat about how best to accomplish this, let me know. 



Tue, 2020-05-05 20:36

Hi Jared,

That would be great. The antigen contains a single N-linked glycan that is close to the CDR-L2 of the antibody. The glycan was built using Coot and the overall model (antigen, glycan, and antibody Fv) was relaxed using FastRelax with the -include_sugars and associated options. Should I optimize the rest of the paratope before trying to optimize the CDR-L2/glycan interaction?


Tue, 2020-05-05 21:12

Hey Chris,

When you build the glycan in Coot, was there decent density?  How big is the glycan?  How is the density for the ends of the glycan?  Are you attempting to optimize affinity with the antibody you have?  


Tue, 2020-05-05 23:02

There was density for the first two sugar residues. I bulit the rest of the glycan based on mass spec data of what the most prevalent glycoform was at that position. The glycan is FA2 and contains 8 residues total. There is no density for the terminal residues. I am trying to increase the affinity of the antibody. 

Wed, 2020-05-06 08:49

Cool. If the glycan is interacting with the antibody in a way that could effect design, I would  model the glycan in the presence of the antibody using the GlycanTreeModeler within Rosettascripts with the option refine="1",  perhaps passing in the density of the antigen to keep the first few sugar residues within the density. Produce a few thousand structures (at least 1k), take the top ones and use those for inputs into RAbD.  You could do this manually, or use ensemble input:

You could, alternatively, design the antibody without the glycan, put the glycan back on ( through the SimpleGlycosylateMover), and use RAbD rosettascript components to optimize the design in the presense of the glycan, perhaps using the GenericMonteCarlo to model/seq design the chosen CDRs. Since you would be essentially starting from a non-modeled glycan here, you would want the density to guide GTM (which works very very well - I really wish I had a preprint to share at this point).

Another thing to think about is to model the glycan with the antibody removed to see some low energy conformations of the glycan apo as well.  The lower affinity of the antibody could be a result of the glycan blocking the antibody from binding.  Do you have K on and K off for this antibody?  

 If you use a newer version of Rosetta (one of the most recent weeklies), all the GTM and glycan defaults have been updated to reflect benchmarking, so you don't really need to much around with options for those. Hope this helps!

Wed, 2020-05-06 11:37