ADL: question about Vina flexible residues and pockets
trott at scripps.edu
Wed Jan 22 13:18:07 PST 2014
If you have known binders for your pocket, as I thought you did, a
retrospective VS is what one would do to evaluate a VS protocol.
If you don't, perhaps you can extrapolate the results from homologous
proteins with known binders (but I guess it comes down to user's
intuition and experience, as I don't know of any quantitative models
If, in the worst case, you have no known binders for your pocket, and
no relevant information from homologous proteins, I don't think there
is much you can do to predict if a VS will turn out to be useful for
your system: perhaps it will work brilliantly, or perhaps it will have
< 1 enrichment (worse than random guessing).
The conventional wisdom is that the "baseline" for good binders would
vary depending on the pocket.
On Wed, Jan 22, 2014 at 10:39 AM, Dennis N Bromley <dbromley at uw.edu> wrote:
> Thank you, Oleg, this is useful. I appreciate your time.
> So far as I know, there is only one known binder for my system and that is
> in a completely different pocket, so I don't really have a set of
> comparables for this protein. (That known binder docks at -6.5 kcal/mol, so
> I was interpreting that as a baseline.) I was basically planning to use
> Vina as a high-throughput pipeline to test lots of different ZINC ligands
> and see if I could zero in on good ones. My problem is that every ligand I
> pick seems 'good', at least by the baselines that I have found.
> So, I apologize if I am not understanding you. Do you see a problem in my
> previous approach or assumptions? Or are you saying that before I can
> begin my previous approach I need to do some preliminary work that I
> haven't mentioned?
> On Wed, Jan 22, 2014 at 10:21 AM, Oleg Trott <trott at scripps.edu> wrote:
>> ... Incidentally, Vina does really well on the COX-2 target in the
>> Directory of Useful Decoys:
>> Only GLIDE HTVS does somewhat better.
>> On Wed, Jan 22, 2014 at 10:10 AM, Oleg Trott <trott at scripps.edu> wrote:
>> > Hi
>> > I'm assuming that you want to evaluate the suitability of Vina as a
>> > virtual screening method for your system.
>> > To do that, you could do a retrospective VS on known binders and
>> > either known non-binders or random decoys, kind of like what the
>> > Directory of Useful Decoys offers (Vina results linked from the Vina
>> > web site and manual). If it sounds too tedious to do by hand, I think
>> > PyRX can automate it, and make ROC plots for you.
>> > The ease of fishing for a few predicted bad binders seems difficult to
>> > interpret.
>> > "Score_only" of X-ray structures probably shouldn't be used to
>> > estimate the binding affinity: for all we know, there is a 10kcal/mol
>> > better energy 0.1A RMSD next to it. (Also, it's believed that about
>> > 1/3 of all bound ligand X-ray structures are completely wrong)
>> > On Wed, Jan 22, 2014 at 9:24 AM, Dennis N Bromley <dbromley at uw.edu>
>> >> Hi everyone,
>> >> I am having a problem because practically any ligand that I throw at my
>> >> pocket seems to dock with reasonable energy and it seems that that
>> >> not be the case. Here is what I am doing:
>> >> I used a pocket-finding algorithm (epos ball-pass) to find a pocket in
>> >> protein and now I want to use Vina to find a ligand that will bind
>> >> to it. I set my search box to be centered around the pocket plus an
>> >> 5 angstroms in each direction for padding. I then set all of the
>> >> pocket-lining residues to be flexible, convert a huge stack of ZINC
>> >> to pdbqt using OpenBabel, and dock them in. My pocket is ~430 cubic
>> >> angstroms.
>> >> To date, the worst energy I have gotten is -5.9 kcal/mol. The best I
>> >> gotten is -12.1 kcal/mol. For reference, the drug celebrex in xtal holo
>> >> position docks into COX-2 (PDB:3LN1) with -10.6 kcal/mol (using
>> >> --score-only) and various bits of literature show Vina energies of
>> >> and such around -9. So, these values seem like they might actually
>> >> potentially bind.
>> >> The problem is that I can't find a bad docker for the life of me, which
>> >> makes me question my good dockers. My pocket is an internal pocket in
>> >> hydrophobic core of the protein (as is my comparable, Celebrex) so I
>> >> searched ZINC for the most awkward polar molecule I could find and it
>> >> came in at -9 kcal/mol. I can't even match my -5.9 molecule; literally
>> >> random picks still score in the -7 to -8 range.
>> >> Any ideas what I am doing wrong? Am I misinterpreting "good energy"?
>> >> Should I not be using flexible residues? There are about twenty
>> >> in the pocket - does that provide so much freedom that it will adjust
>> >> just about any ligand?
>> >> Thank you! I appreciate any insights - something is amiss here that I
>> >> not seeing.
>> >> Denny
>> >> ________________________________________________
>> >> --- ADL: AutoDock List --- http://autodock.scripps.edu/mailing_list---
>> > --
>> > Oleg Trott, Ph.D. (Columbia University)
>> > Staff Scientist in the Olson Lab
>> > The Scripps Research Institute
>> > http://olegtrott.com
>> Oleg Trott, Ph.D. (Columbia University)
>> Staff Scientist in the Olson Lab
>> The Scripps Research Institute
>> --- ADL: AutoDock List --- http://autodock.scripps.edu/mailing_list ---
> --- ADL: AutoDock List --- http://autodock.scripps.edu/mailing_list ---
Oleg Trott, Ph.D. (Columbia University)
Staff Scientist in the Olson Lab
The Scripps Research Institute
More information about the autodock