[Dock-fans] Fixing protonation state

Francesco Pietra chiendarret at yahoo.com
Fri Nov 2 09:41:36 PDT 2007


--- "John J. Irwin" <jji at cgl.ucsf.edu> wrote:

> Francesco
> 
> Francesco Pietra wrote:
> > Hi John:
> > My question was not restricted to the case in point, and nature is not
> faced by
> > the limitations the computational medicinal chemist is. There are a number
> of
> > dockings in nature - known or elusive - where the ligand is large and
> flexible.
> > My question was, is any chance to identify the right pocket when the ligand
> is
> > large and flexible? By DOCK or any other docking program?
> >   
> Identifying the binding site is of course possible, using mutagenesis,
> crystallography, and all that. We start with docking once we know the
> binding site. Are you asking if docking can be used to rank binding
> sites by the likelihood that one binds a particular ligand? Sure, why
> not, give it a try. 

That is just what I am doing. The experimental results are controversial, so
that I wanted to predict from DOCK. I know that is is particularly difficult to
predict what is not known. However, if not else, this first-time approach to
DOCK will introduce me to docking. How could I build a complex for Amber
otherwise?


>But given the famously high false positive rate of docking, the odds are
>against you.

Is any criterion to detect, or suspect (on the basis of DOCK + Amber) false
positives? 

Thanks
francesco 


> > That computational minimization of large, flexible molecules remains a
> > challenge, we know.
> >
> > Thanks
> > francesco
> >
> > --- "John J. Irwin" <jji at cgl.ucsf.edu> wrote:
> >
> >   
> >> Hi Francesco
> >>
> >> Here are three arguments to try to convince you that big floppy
> >> molecules make poor candidates for docking.
> >>
> >> [I was unaware of the target of that compound, and based my comments
> >> about the unsuitability for docking purely upon the molecule itself. ]
> >>
> >> 1. The number of conformations to be sampled goes up exponentially as a
> >> function of the number of rotatable bonds. Let's assume three
> >> conformations around each bond for the sake of simplicity. Thus 3^7 =
> >> 2187 , 3^10 = 59,049 and  3^17 = 129,140,163. These are the number of
> >> conformations to be sampled, even by very coarse sampling, for molecules
> >> with 7, 10 and 17 rotatable bonds, respectively. That's very coarse,
> >> fix-interval sampling. At 1 second per conformation (and that's
> >> considered fast by the standards of the field) that makes 4.1 years for
> >> 17 rotatable bonds. for one *%^* molecule.
> >>
> >> 2. If you have a molecule like the one suggested by Asdrubal, a
> >> "dumbell" with 19 bonds between the aromatics at the end, then a
> >> movement of just a degree or two around the central bond can result in a
> >> relative shift of the extremities (and thus the overall shape of the
> >> molecule) by 5A or more. This hypersensitivity of shape to central bonds
> >> means that you cannot hope to sample all the "right" conformation given
> >> fixed increment angle sampling, which is all you can afford to do. If
> >> you try to sample more finely, say 10 angles around a bond instead of 3,
> >> then you're stuck with 10^17 conformations. Even being clever and doing
> >> 10 points around central bonds decreasing to 3 points at the extremities
> >> would necessitate sampling an implausibly large number of conformations.
> >>
> >> 3. This molecule has terrible entropy problems. It is common practice to
> >> remove molecules from virtual screening libraries with more than 6
> >> methylene's in a row, for this reason. Large numbers of rotatable bonds
> >> tend to have bioavailability problems too (Veber, J Med Chem, 2002), in
> >> case you are still unconvinced.
> >>
> >> By the way, there are many drugs that break the rules I just laid
> >> outlined. I am not saying these molecules cannot be drugs. I am saying
> >> they are problematic candidates for molecular docking, and very likely,
> >> as *starting points* for ligand discovery.
> >>
> >> I hope this is clear and helpful.
> >>
> >> John
> >>
> >>
> >>
> >> Francesco Pietra wrote:
> >>     
> >>> Is your skepticism only about the precision of docking (with large,
> >>>       
> >> flexible
> >>     
> >>> molecules) or is any evidence that also the region of receptor where
> >>>       
> >> docking is
> >>     
> >>> found may be false? E.g. helix S5 instead of S6 in a pore?
> >>> Thanks
> >>> francesco pietra
> >>>
> >>> --- "John J. Irwin" <jji at cgl.ucsf.edu> wrote:
> >>>
> >>>   
> >>>       
> >>>> Hi Asdrubal
> >>>>
> >>>> That molecule has more rotatable bonds (13 in a row, 17 total, + limited
> >>>> sampling on the 2 amides) than I feel comfortable with. It also has
> >>>> molecular weight > 600. These are two reason why the ZINC upload server
> >>>> just won't take it. I have to draw the line somewhere, and I regret to
> >>>> say you have hit it.
> >>>>
> >>>> I would be skeptical of anyone who docks a molecule having 17 (or even
> >>>> 13) rotatable bonds and gets the crystallographic pose within 2A rmsd.
> >>>> Actually, my skepticism starts at around 8 rotatable bonds, but I keep
> >>>> it mostly to myself till we get past 12. ;-)
> >>>>
> >>>> Almost for sure, you are trying to dock into two distinct pockets, and
> >>>> the long aliphatic chain is a linker between them. My suggestion is
> >>>> therefore to dock each end separately. Shoot for something with 7
> >>>> rotatable bonds, even a bit less if you can. Then you can model in the
> >>>> rest of the linker between them.
> >>>>
> >>>> Good luck, and sorry to disappoint.
> >>>>
> >>>> John
> >>>>
> >>>>
> >>>> burgosgu at ualberta.ca wrote:
> >>>>     
> >>>>         
> >>>>> Hi everybody,
> >>>>>
> >>>>> I need to have a compound correctly protonated to dock it. I tried to  
> >>>>> use the resource of
> >>>>> "upload sets" of ZINC, but it seems that it does not have the required 
> 
> >>>>> charactersitics of
> >>>>> bioavailability and non-toxicity, because it gives me no output as it  
> >>>>> did with other
> >>>>> compounds.
> >>>>>
> >>>>> The smile for this compound is:
> >>>>>
> >>>>>           
> >> C1=C(C=CC2=C1C(=C[N]2)CC(=O)NCCCNCCCCNCCCNC(=O)CC3=C[N]C4=C3C=C(C=C4)Br)Br
> >>     
> >>>>> Any suggestions??
> >>>>>
> >>>>> THANKS,
> >>>>>
> >>>>> Asdrubal
> >>>>>
> >>>>> _______________________________________________
> >>>>> Dock-fans mailing list
> >>>>> Dock-fans at docking.org
> >>>>> http://blur.compbio.ucsf.edu/mailman/listinfo/dock-fans
> >>>>>   
> >>>>>       
> >>>>>           
> >>>> _______________________________________________
> >>>> Dock-fans mailing list
> >>>> Dock-fans at docking.org
> >>>> http://blur.compbio.ucsf.edu/mailman/listinfo/dock-fans
> >>>>
> >>>>     
> >>>>         
> >>> __________________________________________________
> >>> Do You Yahoo!?
> >>> Tired of spam?  Yahoo! Mail has the best spam protection around 
> >>> http://mail.yahoo.com 
> >>>   
> >>>       
> >
> >
> > __________________________________________________
> > Do You Yahoo!?
> > Tired of spam?  Yahoo! Mail has the best spam protection around 
> > http://mail.yahoo.com 
> >   
> 


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 


More information about the Dock-fans mailing list