Forcing a conformation by HADDOCK refinement with guessed restraints


I am trying to gauge whether I am posing a problem appropriately for HADDOCK usage. If the approach I am proposing makes any sense, I would appreciate a couple of pointers on where to start…

First of all, Sec13 is a WD40 beta propeller domain (show in rainbow coloring, blue-to-red, below) that interacts with other proteins via a domain invasion motif (DIM) that emulates a blade and wedges itself very tightly into the empty blade slot presented by Sec13 (shown in magenta). So, Sec13 adopts a 7-blade topology but provides only 6-blades; in doing so it generates a binding pocket for the DIM.


What I would like to do is to force a structural model that has had the DIM in the binding site deleted (such as show below) to “close” into a 6-blade topology from the 7-blade topology that it currently adopts.

I was wondering if it makes sense to use HADDOCK to (1) define distance restraints (e.g. using a structural superposition of a blade with the DIM as a best guess) between the 1st and 6th blade and then (2) use HADDOCK as a means to refine against no experimental data and force the whole structure to rearrange into a 6-bladed beta propeller. If this makes sense:
a. what kind of restraints should I use?
b. what kind of refinement should I use?
c. would it make more sense to “dock” the structure as a single chain / single input molecule into HADDOCK or would it be better to break the structure up into chains that correspond to individual blades and then perform a 6-tuple docking?

Thanks in advance for any consideration!


Hi Stefan

In principle you could do it, but this does require some customization.

If you have a local version of haddock2.2 installed, there is an example to refine a protein against PCS data. Have a look at those settings. In short the recipe would be something like:

  • setup a HADDOCK run with a single molecule
  • in run.cns define your molecule as fully flexible and define the proper segment definition:
{* Number of fully flexible segments for molecule (protein) A            *}
{* Note that current max is 5 (edit the run.cns to add more segments     *}

{===>} nfle_A=1;
  • turn on the secondary structure restraints to maintain the secondary structure:
{* Automatically define backbone dihedral angle restraints from structure? *}
{+ choice: none all alpha alphabeta +}
{===>} ssdihed=alphabeta;

depending on what happens you might want to try all instead of alphabeta

  • If you see unfolding, then define some distance restraints to maintain the fold of each domain.

  • Set the number of models for it0/it1/water to a same number, e.g. 200.

  • Set ntrials for it0 to 1

  • And then play with the flexible refinement protocol settings. You might have to increase the number of steps to allow the system to close and fulfil the restraints. Also depending on what happens to your structure, you might want to decrease the temperature in the simulated annealing protocol, e.g.

{===>} tadhigh_t=1000;

{* initial temperature for rigid body first TAD cooling step *}
{===>} tadinit1_t=1000;

{* final temperature after first cooling step *}
{===>} tadfinal1_t=300;

{* initial temperature for second TAD cooling step with flexible side-chain at the inferface *}
{===>} tadinit2_t=500;

{* finale temperature after second cooling step *}
{===>} tadfinal2_t=50;

{* initial temperature for third TAD cooling step with fully flexible interface *}
{===>} tadinit3_t=500;

{* finale temperature after third cooling step *}
{===>} tadfinal3_t=50;

But these are just indicative values. Again, check if your structure unfolds or is stable.

  • Also set noecv to false.

Let us know if you succeed.
Good luck!