ADL: Getting very similar docking results for multiple unrelated proteins (Vina)

HOUSTON Douglas DouglasR.Houston at ed.ac.uk
Wed Aug 19 04:49:56 PDT 2020


Yes, in general, real drugs tend to bind more tightly if they are bigger - up to a point:

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC17830/


In general, docking scores are supposed to correlate with real binding affinity - but the correlation is not all that strong. It's information, but information with a very large margin of error.

________________________________
From: autodock-bounces at scripps.edu <autodock-bounces at scripps.edu> on behalf of Rick Sheridan <ricksher at gmail.com>
Sent: 19 August 2020 12:36
To: autodock at scripps.edu <autodock at scripps.edu>
Subject: Re: ADL: Getting very similar docking results for multiple unrelated proteins (Vina)

I would have to wonder if there isn't more nuance underlying the real world
vs artifact question.  Would be great to have additional commenters
weighing in.  It strikes me that the question one would want to answer is,
"To what extent is the ligand an agonist or inhibitor of the target".
Perhaps it is true that, all things being equal, the scoring function
ascribes a larger binding energy to a molecule with larger polar surface
area exposed to a receptor with corresponding geometry and polarity.  But
what does that say about the agonizing / inhibiting ability of the
real-world ligand against that receptor in the real-world?  Is it
non-information? Or is it useful? i.e. if non-information, why would we pay
attention to docking scores at all?

On Thu, Aug 13, 2020 at 1:20 AM Matt K. <5xx7xx at gmail.com> wrote:

> Thanks a lot!
>
> On Sun, Aug 9, 2020 at 9:59 PM Diogo Martins <diogo.stmart at gmail.com>
> wrote:
>
> > Hi, it's an artifact of the scoring function.
> >
> > On Sun, Aug 9, 2020, 6:09 AM Matt K. <5xx7xx at gmail.com> wrote:
> >
> > > Ok, then a follow up question: is this bias purely an artifact of the
> > > docking software and doesn't translate at all to the real world? Or do
> > real
> > > world drugs also generally bind better if they are bigger?
> > > I'm asking because my screening is purely for repurposing already
> > approved
> > > drugs, hence I don't care about drug-likeness, only which compounds
> will
> > > bind better in reality.
> > >
> > > On Sat, Aug 8, 2020 at 8:47 PM Diogo Martins <diogo.stmart at gmail.com>
> > > wrote:
> > >
> > > > Hello,
> > > >
> > > > It's good to be suspicious of docking scores. Larger molecules almost
> > > > always have better scores, maybe some form of normalization by size,
> > such
> > > > as dividing the score by the number of atoms, could make them a
> little
> > > more
> > > > informative.
> > > >
> > > > Best regards,
> > > >
> > > >
> > > >
> > > >
> > > > On Sat, 8 Aug 2020 at 10:39, Matt K. <5xx7xx at gmail.com> wrote:
> > > >
> > > > > Hello everyone,
> > > > > I'm doing a screen using Vina and I got really suspicious when I
> got
> > > very
> > > > > similar results for three different sites on my protein, so I also
> > ran
> > > > four
> > > > > unrelated proteins chosen at random from PDB, and results are also
> > very
> > > > > similar. Most molecules are within the same percentile (or off by
> one
> > > or
> > > > > two percentiles) across all proteins.
> > > > >
> > > > > The molecules used for docking are from the ZINC15 database,
> > > specifically
> > > > > the "world" subset.
> > > https://zinc15.docking.org/substances/subsets/world/
> > > > > I've downloaded them in mol2 format, split into separate files
> using
> > > > > OpenBabel and prepared them using the MGLTools script
> > > > "prepare_ligand4.py"
> > > > > (this one here:
> > > > >
> > > > >
> > > >
> > >
> >
> http://wind.isi.edu/marbles/assets/components/workflow_portal/users/lib/MGLTools/MGLToolsPckgs/AutoDockTools/Utilities24/prepare_ligand4.py
> > > > > )
> > > > > by using the "bonds_hydrogens" repair option.
> > > > >
> > > > > Protein preparation was done manually in Autodock Tools by deleting
> > > > > ligands, deleting water, adding all hydrogens, merging non-polar
> > > > hydrogens,
> > > > > adding Kollman charges, spreading charge deficit over all atoms in
> > > > residue,
> > > > > and choosing as macromolecule.
> > > > >
> > > > > Grid box for each protein was chosen manually with size usually
> > > 30x30x30
> > > > > angstrom set in each protein's primary binding site. Vina
> > > exhaustiveness
> > > > > was set to 12.
> > > > >
> > > > > What am I doing wrong here? Why are my results so similar across
> > > > unrelated
> > > > > proteins?
> > > > > ________________________________________________
> > > > > --- ADL: AutoDock List  ---
> http://autodock.scripps.edu/mailing_list
> > > ---
> > > > >
> > > > >
> > > > ________________________________________________
> > > > --- ADL: AutoDock List  --- http://autodock.scripps.edu/mailing_list
> > ---
> > > >
> > > >
> > > ________________________________________________
> > > --- ADL: AutoDock List  --- http://autodock.scripps.edu/mailing_list
> ---
> > >
> > >
> > ________________________________________________
> > --- ADL: AutoDock List  --- http://autodock.scripps.edu/mailing_list ---
> >
> >
> ________________________________________________
> --- ADL: AutoDock List  --- http://autodock.scripps.edu/mailing_list ---
>
>

--
--
********************************************************************
Rick Sheridan
EMSKE Phytochem
Head of R&D
Whatsapp:  +1-551-285-8136
www.linkedin.com/in/ricksheridan<http://www.linkedin.com/in/ricksheridan>
<http://www.linkedin.com/pub/rick-sheridan/9/654/515>
Skype:ricksher16
********************************************************************
________________________________________________
--- ADL: AutoDock List  --- http://autodock.scripps.edu/mailing_list ---

The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.


More information about the autodock mailing list