[Dock-fans] RE: [Dockdev] Docking and Scoring algorithms

Vincent LEROUX Vincent.Leroux at edam.uhp-nancy.fr
Sat Feb 26 13:51:20 PST 2005

> I agree that most scoring functions are not correlated enough to the
binding energies. But for me this is not as bad as it seems.

Certainly. All I wanted to say is that in practice there should be
something between MD and docking. My experience was that simply there
is no time to check all the possible hits with MD. It is not only the
CPU time but the file preparation also.

-> The file preparation process can be automated if you have only peptidic ligands (if lucky this is just a matter of using openBabel and writing a Tcl script for NAMD...). At least minimization and interaction energy calculation will give a better ranking of the molecules, if they are close to the correct bound conformations (which should be the case after a successful docking calculation). But after that if there is still too many candidates, dynamics or free energy calculations are too costly...  :-(
I think we agree on that : the gap between flexible docking and MD/FEP is huge in term of precision / computational time.
For the "bridge" that is needed here this is indeed a very difficult problem. The "easiest" way seems to be the design of better scoring functions for the docking programs, that would be almost as precise as calculations using MD programs. Maybe this can be done without them, by using TINKER routines (maybe I will try this one day). But to lower significantly the number of candidates I would prefer more precise docking methods, and particularly by taking care of solvent effects.
At the moment, for the 30k candidates problem, I would try to use GOLD, and if it is not able to restrain further the number of molecules, then there are several possibilities : 1/ the receptor structure is not correct (maybe some MD to try to enhance it? and the inclusion of some water molecules?) 2/ there is no particularly good binder amongst the candidates 3/ GOLD did not understand all the atoms, replaced them by dummies but did the calculation anyway (sad but true, this happens quite often) 4/ docking won't help in this case  :-(

In case of few hundreds of
possible leads the chemists will say finally: "OK, but sort these
somehow then we can start to work with the most promising ones".
Sometimes you must give an order in a week when project deadlines and
milestones are looming.

-> Well then use the docking scores, you have no choice...

Unfortunatelly I have no chance to try SILVER but you made me curious.

-> It is bundled with the latest version of GOLD. You can ask the CCDC for an evaluation version.

Anyway it would be silly to rely on docking only and I doubt anybody
is doing it so.

-> Some experimental chemists ask for docking results and then rely on them... And some weeks later they come and say that your docking results were "bad" because 95% of the hits did not work... Then you reply that 5% are good, are there new binders amongst them? But they are still a bit angry because they assume that you could have identified only those ones solely with docking calculations, and that you made them work too much because you were lazy. Better be very cautious sometimes...

I am not so pessimistic about docking since I saw few times that it
can really find good, specific new binders. My the problem arises not
at the "hit identification" but the "lead selection" and "lead
optimisation" phase of drug development.

> And if computer simulations were faster and more reliable than
experiments, wouldn't 
> many of them lose their job? I am not sure that is what they need  ;-)

Well, put it like this: "both experimental results and simulations are
rubbish" to keep everybody unhappy :)

-> Well one day I was asked to test some new molecules on a target using docking. No one docked well. Then the chemists were happy because they tested them before and they did not work. So it was like "our molecules are crappy, but at least the guy with the computer agree with us on that"...

Motto: "When a new theory is
presented, no one believes it except the theorist. When new data is
presented, everyone believes it, except the person who prepared the
data." --- T. von Karman

-> That's true  :-)


	Vincent LEROUX	PhD student

Equipe de Dynamique des Assemblages Membranaires,
UMR-CNRS 7565, Université Henri Poincaré, Nancy I,
BP 239, 54506 Vandoeuvre-les-Nancy, France

e-mail:	Vincent.Leroux at edam.uhp-nancy.fr

More information about the Dock-fans mailing list