ADL: Negative correlation in AutoDock and Vina calculations - how can I improve it?

Oleg Trott trott at
Thu Dec 12 12:59:02 PST 2013


Why do you give a range of correlations for a single protein?
Shouldn't that just be a number? How many ligands do you use in
calculating each of those numbers?

While you can try to improve your docking results by

(a) making sure you are not making any mistakes in the input
preparation or in using the program (See the Vina FAQ),

(b) perhaps post-process the results e.g. in NNScore (See "Other
Software" in the Vina Manual), and

(c) visual inspection (The users sometimes have insights about their
targets that docking software does not capture),

the end result may still vary, and there are no guarantees (See the
FAQ about accuracy).

Good luck,

On Wed, Dec 11, 2013 at 6:01 AM, Lukas Migas <munieq1111 at> wrote:
> Hello!I recently discovered that some of the results that I've obtained using AutoDock and Vina gave me a negative correlation. The binding modes look promising, but the Docking Scores, when compared to the experimental results give rather poor correlation. I use Spearman test to determine the correlation between the results, and the Spearman correlation often is between -0.7 to -1.I've encountered this problem for two different proteins that I use in docking calculations. However, on another occasion with completely different protein to the two mentioned above, I get excellent correlation which ranges between 0.6 to 0.9.Do you happen to know what can influence this behaviour and how can I improve my results to get better correlation with the experimental results? I am planning to test Ligand Efficiency calculations to determine whether that helps.Thank you for your help in advance,Lukas Migas
> ________________________________________________
> --- ADL: AutoDock List  --- ---

Oleg Trott, Ph.D. (Columbia University)

Staff Scientist in the Olson Lab
The Scripps Research Institute

More information about the autodock mailing list