You are here

The usage of antibody_H3.linuxgccrelease

3 posts / 0 new
Last post
The usage of antibody_H3.linuxgccrelease

Dear all,

I am trying to use the antibody utinity to modeling a camelid antibody (vhh). I used the antibody.linuxgccrelease to produce 10 relaxed models for this  vhh sequence. Then I moved to do the H3 modeling. The "antibody_H3.linuxgccrelease --help" command shows many options for the antibody_H3.linuxgccrelease. However, I am a bit confused about the antibody_H3.linuxgccrelease options.

(1) The"-camelid" and the "-camelid_constraints" options. I think I should add this "-camelid" option in the "flags" file, but actually the program can run without this option. So is it neccessary when working with camelid antibody? And is the "the "camelid_constraints" option neccessary? when should it be used? 

(2)  The "-l" and the "-nstruct" options. I want to do H3_modeling for the 10 models generated by antibody.linuxgccrelease, so I used "ls grafting/*.relaxed.pdb > grafted_models.list" to get a list, which looks like:


And I included the "-l grafted_models.list" option in the "flags" file.  According to the disccussion in the forum, I learned that at least 1000 modeled need to be produced in the H3_modeling for one initial template. So for the "-nstruct" option, should I pass 1000? Since there are 10 initial models, should the value be 10,000? 

(3)  The "antibody:constrain_cter" and "constraints:cst_weight" options. I guess these two options are both neccessary for vh-vl and vhh antibodies, aren't them? How should the value for the "constraints:cst_weight" option be determined?

(4) The "antibody:auto_generate_kink_constraint" and "antibody:all_atom_mode_kink_constraint" options. Are they necessary for the vhh?

(5) The "antibody:constrain_vlvh_qq" option. This option should not be included for the camelid antibody, should it?

Thank you very much!

Post Situation: 
Thu, 2020-05-28 22:50


  1. The camelid antibody should be auto detected, and you do not need to add these flags. We will actually deprecate them, so thank you for pointing this out.
  2. For your camelid antibody, it is not necessary to produce 10 grafted models. The 10 grafted models are recommended for heavy/light Fv domains to capture different orientations of the two domains relative to each other. Since you only have one domain, this is not necessary. You can proceed by just using one of your relaxed models and the “-s” option instead of “-l”. They should all be very similar to each other. You can then use -nstruct 1000 to get a total of 1,000 models.
  3. The constraints you mention are both recommended to use for all antibodies, including VHH antibodies. So you’d add the -antibody:constrain_cter option and the -constraints:cst_weight 1.0
  4. The kink constraint, constrains the CDR H3 loop into a “kinked” conformation that is commonly observed in antibody structures. It’s recommended to use this constraint, so you should add both flags.
  5. The antibody:constrain_vlvh_qq constraint constrains the distance between a glutamine in the light chain and a glutamine in the heavy chain, that frequently form hydrogen bonds with each other. So you do not need this option for a camelid antibody.


In the most recent versions of Rosetta (after May 17th, 2020) the constraints mentioned in 3. and 4.  (and 5. ) are turned on by default and you no longer need to provide the flags unless you specifically want to turn the constraints off. In older versions of Rosetta you need to specify them as described above. I hope this helps!

Fri, 2020-05-29 12:42

Thank you rahelf,

These are very helpful!

Thu, 2020-06-04 08:20