The Molecular Graphics Laboratory Forum

AutoDock, AutoLigand, MGLTools, Vina, PyRx and more.
It is currently Tue Jul 17, 2018 2:09 am

All times are UTC




Post new topic Reply to topic  [ 5 posts ] 
Author Message
PostPosted: Wed Jul 15, 2009 11:44 am 
Offline
Micromolar User
Micromolar User
User avatar

Joined: Thu Jul 09, 2009 8:59 am
Posts: 17
It more and more appears that Arieh Warshel is right for insisting that enzymes achieve their catalytic power mainly by specific electrostatic transition state stabilization, i.e. lowering the activation barrier by intercepting-stabilizing the transient-sudden changes of local charges at the ultra-short-living transition state (and the energetic cost for being able to do this was paid for by the folding of the enzyme). A seminal reference is "Electrostatic basis for enzyme catalysis" by A. Warshel et al, Chem. Rev. 106, 3210-3235 (2006). Good enzyme inhibitors are then often shape (including H-bonds and hydrophobics) as well as local-charge mimics of the transition state. Which leads to the question: since Vina does not use local charges (at least not directly), may one then fear that virtual screens with Vina on enzyme-receptors can miss important local-charge-analog leads?


Top
 Profile  
 
PostPosted: Wed Jul 15, 2009 7:41 pm 
Offline
MGL member
MGL member
User avatar

Joined: Mon Oct 20, 2008 9:39 pm
Posts: 202
wim nerinckx wrote:
Which leads to the question: since Vina does not use local charges (at least not directly), may one then fear that virtual screens with Vina on enzyme-receptors can miss important local-charge-analog leads?


Wim,

I'm not especially worried about missing leads, i.e. false negatives, as nonchalant as it sounds. It's easy to see that if you are doing virtual screening and choosing the top 1% for further testing, then a 1% false positive rate is as bad as a 50% false negative rate. Changes to scoring that decrease false negatives at the expense of increasing false positives may not necessarily be good.

Secondly, false negatives are probably very difficult to avoid anyway (induced fit, etc.), but again, that's OK.

Thirdly, Vina's approach to scoring functions is basically "machine learning", i.e. it uses what appears to work, not what should have worked in theory. AD4 uses Coulomb-like terms, and I also experimented with them, but the experiments didn't appear promising.

That said, we haven't tried to specialize the scoring function for enzymes or any other class. And in the spirit of this approach, if we find new informative features of the receptor-ligand interaction, we will certainly consider adopting them.

I could theorize about why one approach works better than another, but this would basically be an appeal to intuition rather than rigorous science, and not the real reason why one approach was chosen over another (Perhaps poor charge assignment creates more noise than signal, or perhaps electrostatic fields tend to be better compensated in reality than would be expected from naive models).

_________________
If you don't get your question answered here, consider posting it on the AutoDock mailing list instead of this forum. Please do not email or PM me with requests for help.


Top
 Profile  
 
PostPosted: Mon Jul 20, 2009 1:18 pm 
Offline
Millimolar User
Millimolar User
User avatar

Joined: Fri Jul 17, 2009 7:30 am
Posts: 2
I am performing an “in-house” limited validation of your very interesting algorithm, and I do completely agree with you that results are more important than “what should have worked in theory”. So, if Coulomb-like terms do not improve results, it is ok for me to ignore charges.
However, I still have a problem dealing with charged structures.
The number of “polar” hydrogens in a ligand structure depend on the charge of the ligand. Just think to carboxylic acids and carboxylate anions. In this case, are we supposed to submit the “un-charged” structure or stick with the chemically more reasonable charged structure? (in the protein, hydrogens in the default ADT structure are consistent with the “charged” structure)
I tried an example were both binding site and ligand are strongly charged: neuraminidase and a couple of well known inhibitors. In my hands, Vina was unable to simulate the correct binding and the binding energy (I hope I did not do some stupid error. Pertinent files are attached)
In conclusion: In my (limited) “validation set” I do obtain extremely good results with low-polarity binding sites and ligands (mainly kinase inhibitors), but unsatisfactory dockings with polar systems. Am I doing something wrong?

(moderator: please post in the thread you are following up to)


Attachments:
File comment: output
LigCharged_out.pdbqt [27.59 KiB]
Downloaded 262 times
File comment: Ligand (charged)
LigCharged.pdbqt [3 KiB]
Downloaded 272 times
File comment: Experimental pose of the ligand
1F8C_LigExp.pdb [1.79 KiB]
Downloaded 288 times
Top
 Profile  
 
PostPosted: Tue Jul 21, 2009 4:44 pm 
Offline
MGL member
MGL member
User avatar

Joined: Mon Oct 20, 2008 9:39 pm
Posts: 202
I don't know if it's a coincidence or something else, but I'm also working with neuraminidase. I'm getting pretty good results so far...

_________________
If you don't get your question answered here, consider posting it on the AutoDock mailing list instead of this forum. Please do not email or PM me with requests for help.


Top
 Profile  
 
PostPosted: Thu Jul 30, 2009 8:33 am 
Offline
Millimolar User
Millimolar User
User avatar

Joined: Fri Jul 17, 2009 7:30 am
Posts: 2
Of course I did something wrong. I forgot the flexible rings problem in ADT (see the Autodock 4.2 user guide, appendix II). Results with neuraminidase inhibitors are actually quite good (at least, as far as RMS is concerned). By the way, it would be a nice extension in VINA an algorithm dealing with macrocyclic rings conformation.


Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 5 posts ] 

All times are UTC


Who is online

Users browsing this forum: No registered users and 6 guests


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
POWERED_BY
Translated by Maël Soucaze © 2009 phpBB.fr