ADL: Solvation Parameters

zgleo@tom.com zgleo at tom.com
Sun Sep 12 05:01:22 PDT 2004


Dear Garrett M. Morris£¬
     The setting of "rmstol" doesn't work in fact. Whatever value you change the setting to, autodock always uses 1 Angstrom as the real tolerance RMSD. Maybe it's a small bug occurring during the read of the dpf file.

>This is a very good question.
>
>You can be pretty certain if you obtain just one cluster at the end of 
>the docking; 10 dockings per compound may be enough, but you are doing 
>very "deep" dockings, at 50 million evals.  You might like to try 100 
>dockings at 5 million evals, to compare the how best to "spend" the 
>same number of energy evaluations ("evals").  See if you get reasonable 
>docked conformations at the end of each kind of docking; 100 dockings 
>is more likely to give you better clustering statistics.  If you start 
>with random number generator seeds that are dependent on the time and 
>process id (pid) of your docking (this is the deafult behaviour for 
>AutoDock scripts like mkdpf3 and ADT), then you should get different 
>results each time: remember that AutoDock's search methods, SA, GA and 
>LGA, are all stochastic search methods, which means the outcome will 
>vary depending on the random numbers generated during the docking.
>
>You can be confident of convergence if you obtain the same 
>conformations frequently or even always, and this will be reflected in 
>the conformational clustering in the histogram at the end of the DLG 
>(docking log file).  The clustering depends on the size of your ligand 
>and the rmsd tolerance you use, but it is best to use an "rmstol" of  
>about 2 Angsroms for your clustering.
>
>I hope this helps,
>
>Garrett


Best Regard!
                                 Zhiguo Liu
                                 DDDC
                                 Chinese Academy of Science
                                 Shanghai,China
                                 zgleo at tom.com
                                 2004-09-12





More information about the autodock mailing list