1

I have a PDB structure that I'm analyzing which has a putative binding pocket in it for some endogenous ligand.

I'd like to to determine for the relevant amino acids within, say, 3A of the pocket, where the optimal hydrogen bond donor/acceptor pair for the ligand would be within the pocket. I have done this already for determining locations of optimal pi-pi stacking (e.g. find aromatic residues, determine plane the face of the ring, go out N angstroms orthogonal to the face), but i'm struggling to consider this for hydrogen bonds.

Can this be done?

Raymond Chen
  • 44,448
  • 11
  • 96
  • 135
DrTchocky
  • 515
  • 1
  • 5
  • 14
  • do you consider your pocket rigid/fixed, or do you need to evaluate all the possible/hypothetical pocket avalable conformations ? – pippo1980 Jun 20 '22 at 09:36
  • @DrTchochy, out of curiosity how a pocket is described in PyMol/Biopython ? – pippo1980 Jun 20 '22 at 16:34
  • 1
    @pippo1980 For now, I just want to consider a rigid pocket. Currently, the pocket is defined a number of ways, the most accessible likely being an array of 3d grid points (x/y/z coordinates) that define the space. I can easily define it a number of other ways if it makes the problem easier to approximate/"solve" – DrTchocky Jun 21 '22 at 16:48
  • Dont have a solution for u just intrigued by the question. Thanks for explanation about the pocket, yesterday was googling about how are pockets described by different SW, had difficulties to understand the logic of PyMol PyVol https://schlessinger-lab.github.io/pyvol/index.html . So your pocket is not a real pocket but an approximation of it? how fine is your grid? How does pdb resolution limit relate to grid spacing (i.e. for a 2.3A resolution pdb it makes no sense to have a pocket described by a grid with XX A spacing ) ? – pippo1980 Jun 21 '22 at 17:01
  • Not sure my questions are legitimate/make sense , but I am trying to figure out if I am getting the question right – pippo1980 Jun 21 '22 at 17:01
  • tried to figure out a way to answer your question. Does this seem reasonable to You or looks flawed by some wrong assumptions oe mistakes ? Let me know, need to learn too. – pippo1980 Jun 22 '22 at 15:59
  • oh forgot to read the : "...I have a PDB structure that I'm analyzing which has a putative binding pocket in it for some endogenous ligand...", why dont you just run a couple of modeling run (autodock, or gold if you have them) and see where the results point too. Not an expert in docking, but think the take home message is repeat docking a lot of time , possibly with different aproaches and go where the more representative result goes. Then do mutagenesis, kinetics of mutant, and hope for a co-crystal – pippo1980 Jun 22 '22 at 16:17

1 Answers1

1

Weel, I'll try to write out how I would try to do it.

First of all its not clear to me if your pocket is described by a grid that represent the pocket surface, or by a grid that represent all the pocket space (lets call it pocket cloud).

With Biopython assuming you have a cloud described by your grid:

Loop over all the cloud-grid points:
 
       for every point loop over all the PDB atoms that are H donor or acceptor:

           if the distance is in the desidered target range (3A - distance for optimal    
                                                    donor or acceptor pair):
     
               select the corresponding AA/atom/point

               add to your result list the point as donor/acceptor/or both togeher 
                                                          with the atom/AA selected

           else:
 
               pass

with Biopyton and distances see here: Biopython PDB: calculate distance between an atom and a point

H bonds are generally 2.7 to 3.3 Å

I am not sure my logic is correct, the idea is to end up with a subset of your grids point where you have red grid points where you could pose a donor and blue ones where you could pose an acceptor.

We are talking only about distances here, if you introduce geometry factors of the bond I think you should need a ligand with its own geometry too

Of course with this approach you would waste a lot of time on not productive computation, if You find a way to select only the grid surface point you could select a subset of PDB atoms that are close to the surface (3A) and then use the same approach above.

pippo1980
  • 2,181
  • 3
  • 14
  • 30