ADL: Vina: docking results dependent on ligand input conformation?

Oleg Trott trott at scripps.edu
Tue Apr 15 14:30:47 PDT 2014


Hi

> What kind of "techniques" could lead to better results with crystal ligand coordinates?

These are often subtle and mathematical flaws that may require a long
explanation and still are easily misunderstood, so I don't think I
want to go into details. Essentially, they are bugs, but whereas the
usual bugs might make the program crash, etc., these let it "cheat"
and on the redocking test. This is one of the reasons they are so
common: When the authors introduce the bug unwittingly, they think
they improved the algorithm, and they keep it, disseminate the
results, etc.

Oleg

On Tue, Apr 15, 2014 at 2:27 AM, liu junjun <ljjlp03 at gmail.com> wrote:
> Hello Oleg,
>
> I have similar question on this issue. Please help me if you have time.
> Thanks in advance!
>
> On page 13 of Auttodock's usergide version 4.2.5, there is a statement when
> talking about the radomize command:
> "The techniques used to implement rotation of the ligand may be biased to
> prefer values of zero rotation"
> The guide thus recommends user to randomize starting coordinates when
> redocking.
>
> I also read from one or more papers on the performance of Autodock (sorry I
> don't remember which paper) that the hit/success rate for redocking with
> input ligand from its crystal structure with protein is much higher than
> those from randomly generated conformations. What I understand is that
> results from Autodock bias the intial ligand coordinates.
>
> You mentioned that vina is unbiased the input. Could you please also
> comment on the bias behavior of autodock? What kind of "techniques" could
> lead to better results with crystal ligand coordinates? Is it necessary to
> optimized the ligand before docking?
>
> Thanks a lot!
>
> Junjun
>
>
> On Tue, Apr 15, 2014 at 11:27 AM, Oleg Trott <trott at scripps.edu> wrote:
>
>> Hi
>>
>> If there are two low-energy basins separated by a high-energy barrier,
>> AD4 and Vina should sample both basins, given enough
>> time/exhaustiveness, not unlike the fact that a coin toss should give
>> you "tails" given enough samples.
>>
>> Furthermore, for Vina, as a consequence of its property of being
>> unbiased, the probability of finding your results in one or the other
>> basin should not depend on the input conformation (i.e. the values of
>> the ROTATABLE torsion angles, position and orientation in the input
>> structure). I'm not sure the same holds for AD4 -- at least in the
>> past it didn't.
>>
>> The issue you bring up is a separate one: in Vina and AD4, torsions
>> are either totally rigid, or freely rotatable (unfortunately), so it's
>> not an issue of Vina staying in a certain basin because of a barrier,
>> but of the existence of the other basin, as far as the docking program
>> is concerned.
>>
>> So, if you want to model a molecule where a torsion is expected to
>> only take values near, say, 0 or 180 degrees, you would probably need
>> to dock the two conformations (with the torsion "frozen") separately,
>> as you suggest. Another approach is to let such torsions be rotatable,
>> but filter out the implausible results in, say, a virtual screen.
>>
>> Oleg
>>
>>
>>
>>
>> On Mon, Apr 14, 2014 at 5:00 PM, Soren Wacker <swacker at ucalgary.ca> wrote:
>> > However, what if there are different conformers that are separated by a
>> big energy barrier? Autodock and Vina will stay in a local minimum and not
>> sample the full conformation space, right? E.g. Cis- and trans
>> conformations in 1,3-dimethyl-urea. This actually happens because amide
>> bonds are considered rigid by default. In such cases it makes totally sense
>> to perform the docking with the different conformations and the output will
>> depend on the particular conformation of the input molecule. At least, if
>> the conformers have similar chemical stability.
>> >
>> > Would you agree?
>> >
>> >
>> > --
>> > Dr. Soeren Wacker
>> > Centre For Molecular Simulation, BI # 547
>> > The University of Calgary
>> > 2500 University Drive (NW)
>> > Calgary, Alberta
>> > Canada, T2N 1N4
>> >
>> > Tel.:   ++1 (403) 210 7860
>> > Email:  swacker at ucalgary.ca
>> >
>> > ________________________________________
>> > From: autodock-bounces at scripps.edu [autodock-bounces at scripps.edu] on
>> behalf of Oleg Trott [trott at scripps.edu]
>> > Sent: Monday, April 14, 2014 10:44 AM
>> > To: autodock at scripps.edu
>> > Subject: Re: ADL: Vina: docking results dependent on ligand input
>> conformation?
>> >
>> > Hi
>> >
>> > The probability distribution of the results coming from Vina should
>> > not depend on the input ligand's conformation (position, orientation
>> > and torsions). In other words, its results should be unbiased w.r.t.
>> > the input. (This is a quality many, if not most, docking programs
>> > lack)
>> >
>> > This is not quite the same as saying that the individual results do
>> > not depend on the input, because they are merely samples from the
>> > above distribution. These samples will depend on the random seed, and
>> > similarly, they may be affected by any changes in the input.
>> >
>> > Oleg
>> >
>> > On Mon, Apr 14, 2014 at 5:03 AM, Juan Munoz-Garcia
>> > <juan.munoz-garcia at bioch.ox.ac.uk> wrote:
>> >> Hi,
>> >>
>> >> could any of the experts of Autodock Vina please tell me if they’ve
>> seen any dependencies of docking results on the input conformation of the
>> ligand? I’ve seen this with other scoring functions, specially for known
>> low affinity ligands, so that to minimise this effect I submitted and
>> ensemble of randomly generated conformations of each ligand. Would you
>> suggest to do the same with Vina?
>> >>
>> >> Thank you.
>> >> Regards.
>> >>
>> >> Juan C. Munoz-Garcia, PhD
>> >> Biomembrane Structure Unit
>> >> University of Oxford
>> >> South Parks Road
>> >> Oxford OX1 3QU
>> >> juan.munoz-garcia at bioch.ox.ac.uk
>> >>
>> >>
>> >>
>> >>
>> >> ________________________________________________
>> >> --- 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
>> >
>> > ________________________________________________
>> > --- 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
>>
>> http://olegtrott.com
>>
>> ________________________________________________
>> --- 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

http://olegtrott.com



More information about the autodock mailing list