The Molecular Graphics Laboratory Forum

Input for future features of AutoDock, Vina & ADT
Page 1 of 2

Author:  oleg_trott [ Fri Jan 28, 2011 6:59 pm ]
Post subject:  Input for future features of AutoDock, Vina & ADT

Dear AutoDockers,

We are in the process of preparing a grant proposal to the National Institutes of Health to help fund the continuing development, maintenance and user support of AutoDock, AutoDock Vina, AutoDockTools and related virtual screening software.If you or your colleagues have suggestions for improvements or new features to help you with your docking and virtual screening research we would like to hear from you.You can post them on this listserver, on our forum (, or if you prefer e-mail me directly (

Thank you for your continued interest and support of the AutoDock suite of programs.

Happy Docking,

Art Olson

Please post your suggestions as a reply to this thread, or send them by email to the mailing list above or to Prof. Olson. (We will remove this topic from "announcements" in about 30 days)


Author:  wim nerinckx [ Fri Jan 28, 2011 7:37 pm ]
Post subject:  Re: Input for future features of AutoDock, Vina & ADT

Dear AutoDock / ADT / Vina developers,

Although already often remarkably predictive as these docking programs are now, both calculate ligand-into-receptor poses on the basis of interaction fields generated from centres of atoms (please excuse me for this oversimplification). However, certainly for many heteroatoms it is not the centre of the atom that is the attractor or repellant. E.g. an ether-oxygen of a ligand has two lone pairs each occupying their own space (and doing their own interactions) away from the oxygen's centre. Same for an oxygen at a keto-function (e.g. in all the backbone-peptide bonds), etc... Analogous for the orbitals of multi-coordinating metal ions.

What if for such heteroatoms "excentric interactors" could be introduced? E.g. oxygens would then have two excentric interactor points for the grid-calculation and for the docking, instead of only one at their centre (each "excentre-point" then being near the centre of the respective lone pairs). Idem for metal ions which would then have several excentric interaction points that may be able to mimick their coordination towards the receptor and/or the ligand. I realize that this may be difficult and drastic to implement since it may require yet another 3D-coordinate format, e.g. ".pdbqte" with "e" for "excentric" (or a .pdbqt - .pdbqte interconversion could be built-in during the program run), but it may have a chance to further improve the docking predictions.

Big kudos for your great work on these marvellous programs,
Wim Nerinckx, PhD
Ghent University, Belgium

Author:  akukol [ Sun Jan 30, 2011 11:37 am ]
Post subject:  Re: Input for future features of AutoDock, Vina & ADT


A recent publication suggested (Garcia-Sosa et al., J. Comput. Chem. 2009) that the correlation of predicted ligand binding energies with experiments can be substantially improved by simple measures such as dividing the AutoDock binding energy by the water/butanol partition coeffcient for that molecule or the number of carbon atoms. Such simple measures are not very costly in terms of computation time. Maybe some of these measures could be included in the ligand-receptor interaction energy evaluation during the optimisation.

In my own studies with the directory of useful decoys I applied applied a rescoring of the binding energies ADscore: log10 (-ADscore/P), where P is the XLOGP3 predicted water/butanol partition coefficient. However, I found improvement in a few cases with AD4, while it got worse in other cases. On average over 10 targets, there was only a slight improvement of results. But maybe there is scope for improvement here.
The results I looked for were the Receiver Operator Curve Enrichment (ROCE) in a virtual screening experiment at 2% of the screened database of known ligands/decoys: ROCE 2%

The same method applied to Vina made results worse every time.

Many thanks for providing that great software
University of Hertfordshire, UK

Author:  bbennion [ Thu Feb 10, 2011 12:32 am ]
Post subject:  Re: Input for future features of AutoDock, Vina & ADT

PyRx feature. Proper coloration of other halogens besides flourine.

Author:  akukol [ Mon Feb 14, 2011 8:57 am ]
Post subject:  Re: Input for future features of AutoDock, Vina & ADT

The first step for docking in AutoDock is the calculation of atom affinity maps on the receptor for each atom type of the ligand with AutoGrid (not sure if Vina does the same, probably internally).

What about using fragments of functional groups rather than atoms to calculate the affinity maps ? Or the most accurate approach might be to get rid of affinity maps/autogrid altogether and evaluate the affinity with the receptor for each ligand individually. Problems with this approach might be the that the calculation time is prohibitive, and if and how ligand conformations should be taken into account. So the use of functional groups might be a better approach.

University of Hertfordshire, UK

Author:  akukol [ Wed Feb 16, 2011 12:45 pm ]
Post subject:  Re: Input for future features of AutoDock, Vina & ADT

The parameters of the binding energy function of docking programs are usually optimised to reproduce the binding energy of ligands.

Would it be useful to to include non-ligands in this optimisation process ? This may lead to a better discrimination between true ligands and non-binding molecules. It may lead to a worse reproduction of bound ligand conformations, but two alternative sets of parameters could be provided; one for reproducing bound ligand poses and another one for discriminating between ligands and non-ligands.

Most databases such as pdbbind focus on true ligands, but molecules that are confirmed not to bind to a receptor may be found in the ChEMBL database.

University of Hertfordshire, UK

Author:  alesser [ Tue Feb 22, 2011 7:01 pm ]
Post subject:  Suggestion: Swiss Dock ?

While I know this isn't developed at the same place as PyRx, it might be really cool for PyRx to become a multi-docking-program tool to execute a given docking with many different engines for consensus binding modes / energetics.

Along these lines, you might consider the inclusion of SwissDock ( which has initial SOAP access available:

Author:  wim nerinckx [ Fri Feb 25, 2011 10:35 am ]
Post subject:  Re: Input for future features of AutoDock, Vina & ADT

Would it be possible to "reverse-engineer" a gridfile (or a virtual pdbqt-file) for AutoDock --and especially for Vina since it is a much faster program-- based on (a pose of) the structure of an already known inhibiting molecule? That would then yield a "virtual pdbqt" pharmacology model that could be useful for virtual screenings on the vast majority of receptors that don't (yet) have a 3D-structure...

Author:  olson [ Fri Mar 04, 2011 1:50 pm ]
Post subject:  Re: Input for future features of AutoDock, Vina & ADT

Thank you all for the thoughtful suggestions for improving AutoDock, AD Vina and related codes. We have considered them all, and in fact have placed many on our "to do" list. Hopefully with continued funding of our project you will see these improvements in due time.

We are deeply appreciative of the strong support of the AutoDock community, and we encourage the continuing conversation.

With best regards,

Art Olson

Department of Molecular Biology
The Scripps Research Institute
La Jolla, CA 92037

Author:  Douglas Houston [ Tue Mar 08, 2011 4:34 pm ]
Post subject:  Re: Input for future features of AutoDock, Vina & ADT

I already sent a big list to the mailing list, but I've had one more idea.

The introduction of the "gradient" idea as used by Vina to give give directionality to the docking in Autodock. As I understand it, this element of the optimization algorithm in Vina contributes to its (massive) speed advantage over Autodock.

And support for Windows 7 and Radeon cards for MGLTools would be good (see

Page 1 of 2 All times are UTC
Powered by phpBB® Forum Software © phpBB Group