Cryo EM data not constraining structure properly

Dear Dr. Bonvin,

I have been trying to use HADDOCK to dock two molecules to a lipid containing particle. I have cross-linking constraint data as well as a low resolution cryoEM envelope. I have been measuring my centroids in ChimeraX for input and my EM data is in an .mrc file (CCP4 map file). Basically the data that I get back is not centered in the EM density and it does not look like the EM density data was used as a constraint when I manually overlay the output structure to the map. I am wondering if my file format is not compatible or if I am missing a setting (perhaps the default scale settings I am using are not right?). Would love to get this problem solved. The run in question is Davidson_024 that was submitted today. I can send the files and constraints I used if it will help. Thank you.


There could be many reasons for that.

First could it be the density was moved in Chimera, and you did not save that file? I.e. the centroids positions and the density map are not consistent?

Also in case you would have your nanodisk already fitter into the density, you could turn on the option to keep it in its original position

Thanks for the response. I don’t think the density was moved. I have fit it in both Chmera and Powerfit and the centroid coordinates are very similar. I have tried using the coordinates as either ambiguous or unambiguous restraints and am having the same problem either way (I assume that is the option you are talking about?). After reading more of your papers, I saw that HADDOCK starts out with the proteins far apart in space then brings them together during the fitting. I am wondering if the program simply cannot converge on a solution and thus never moves them closer. There are parts of the N- and C-terminus of one of my components that hasn’t been visualized by crystallography. If those are part of the docking interface, It may just not be able to find a solution. Doesn’t explain why the structures are not in the densities after the final “fit” though. I have tried running the dockinq with and without the cross-links, but same problem each way. I have manually developed a model which fits the components into the densities and have minimized the cross-link distances. But I would love to have HADDOCK’s confirmation and to make sure no steric clashes etc. before I write this paper. Any further ideas you have, I am all ears.

Hi Sean

Are the two centroids you are defining corresponding in turn to the nano disk and the other protein?

If that’s the case I would then use unambiguous centroid restraints.

Also if you can pre-fit the nano disk into the density, try saving the coordinates, and when uploading to the server turn on the option to keep the first molecule (nano disk) in its original position.

Fix molecule at its original position during it0? → true

Thanks for your continued help. I appreciate your time and patience with me. Yes, the two centroids correspond to the nanodisk and the other protein (LCAT). I have saved the coordinates after fitting to the density for both the disc and the ligand. When I pull them up in PyMol, they sit right where they are supposed to in the EM density. I have tried keeping the centroids unambiguous by “Are centroid restraints ambiguous?” = false for both. I have also tried that with = true for both. No luck. I have turned the centroids restraints both on and off with the “Use density /XREF restraints” both on or off.

I can’t exactly figure out where you are referring to when you say “Fix molecule at its original position during it0?” I don’t see this option under the EM restraint dropdown. I looked through all the other dropdowns and don’t see that option. I think this would be very helpful if I can find out where that option is located. I have access to the Expert level of the online portal. Do I need to be a “Guru”?


The fix molecule option is in the molecule menu, just below where you upload the pdb files. It is not specific for crying-EM but more generic


That seemed to do the trick! Both molecules are properly fitted into the density now. Gonna play with the parameters more to make sure that I am not over biasing the fit by being too restrictive, but this seems to have solved my problem. Thank you very much for your help.


Did you only fix the 1st one?
To allow more freedom to the smaller component.

You could also consider turning on Center of mass restraints to have a bit more compact solutions as your crosslinks are quite loose.

My initial test fixed both. I am going to try various combinations of taking that down. I have my cross-links limits loose right now because I worried that the program couldn’t find a solution with tighter tolerances. I will drop those down too.


HADDOCK won’t have problems with tighter tolerances. The restraints might be violated, but this will not prevent generating docked solutions.

If you want to check the consistency of your restraints (i.e. are there solutions consisting with all of them), you can try to run our DISVIS server. It will show you the accessible interaction space consistent with your data

and related tutorial at:

Yep. I have made extensive use of DisVis. It is a great tool. I used it in an iterative fashion on 8 experimental cross-links to realize that there are at least two docking poses at two reciprocal sites on the nano disk. This fits with the cryo-EM density. I have a draft of a manuscript on all of this. Once I get the final model figured out it would be great to have your feedback on how I used the tools if you have time.

1 Like

Very nice!