You are here

How to specify constraints during fold-and-dock?

4 posts / 0 new
Last post
How to specify constraints during fold-and-dock?

I'm folding a protein which is a trimeric long helix bundle.  I started from the main/demos/public/symmetry_examples/fold-and-dock/ and added my info (sequence, fragments etc).  This basically works but a lot of the resulting folds are globular not long - the helixes fold over on themselves.  I'd like to constrain this to extended conformations only.

I looked at constraints in the Rosetta docs, and tried creating input_files/constraints.cst to give a penalty to any structure which has less than 50-70 A distance between residues 1 and 63:

AtomPair CA 1A CA 63A FADE 50 1000 20 -100 100
AtomPair CA 1B CA 63B FADE 50 1000 20 -100 100
AtomPair CA 1C CA 63C FADE 50 1000 20 -100 100

I also added this to inputs/fold-and-dock-flags to turn on the constraints:

-constraints:cst_fa_file ./input_files/constraints.cst
-constraints:cst_fa_weight 2.0

However this has no effect; most of the resulting structures are compact, same as before.

How do I turn on constraints for fold-and-dock with symmetry?  Is there a problem with how I'm specifying the constraints?

Many thanks!

(My normal fold-and-dock flags before the constraints are attached; Environment is Rosetta 2020.11 on Ubuntu 18.04 on AWS)

fold-and-dock-flags.txt1.73 KB
Post Situation: 
Tue, 2020-05-12 01:55

EDIT: I did not read your question well enough and thought you were aiming for a range between 50 and 70, but now I notice you mentioned you do want the upper limit huge and was done on purpose. So my message was all wrong. But I would suggest trying BOUNDED. And adding the weight to the centroid scorefunction (cen) in addition to the full-atom (fa) one.

The problem is your bounds I reckon.

  • FADE function
  • 50 lower
  • 1000 upper (!!!)
  • 20 depth/distance
  • in well bonus -100
  • out of well penalty 100


The upper should be 70, while the d is how quickly you get the bonus. That is the function has a pontential well shape, which starts decreasing at lower and reaches full depth at lower + distance and starts going up at upper + distance, to be fully up at upper.

So FADE 45 75 5 -100 100 would be better. That is at 45 Å you pay 2 * 100 kcal/mol, while with 50-70 Å you gain -200 kcal/mol (those are mad high numbers btw!).

I personally unimaginatively would have gone with FLAT_HARMONIC centered at 60 Å ± 5 with tollerance 10 and have not used this, so I may be reading this wrong.

The FADE function is less common than BOUNDED or FLAT_HARMONIC, but has the feature you get a bonus in addition to a penalty. So makes total sense here.




Wed, 2020-05-13 10:56

Thank you very much!  I think you solved it - fold-and-dock uses centroids, so the constraints probably weren't being applied because of that. 

I will try and report back. 

Btw it seems like the centroid constraints are set with -constraints:cst_file and fullatom with -constraints:cst_fa_file?  Does my syntax look okay except for that?  It would be good to add this to the documentation at -  now it doesn't explain about the difference between the two.

FLAT_HARMONIC or BOUNDED would both work - I should probably use a more common type of function.  I wanted to essentially rule out any compact conformations, hence the high numbers :)

Wed, 2020-05-13 17:41

I should mention that for amino acids in centroid the side chain is shrunk down to the beta carbon and a lot of atoms would not be there, here is phenylalanine in centroid:

ATOM  N   Nbb  NH1  -0.47
ATOM  CA  CAbb CT1  0.07
ATOM  C   CObb C    0.51
ATOM  O   OCbb O    -0.51
ATOM  CB  CB    CT2  -0.18
ATOM  H   HNbb H    0.31

So the _fa_ to _cen_ works in your case, but no for other cases.

About the confusingness, it makes a bit more sense with pyrosetta where you explicitly apply the constraints to the pose and set the weight to the scorefunction. And a centroid pose is different than a full atom one.

Thu, 2020-05-14 07:01