ADL: Vina: 3 quick and simple questions

Rentzsch, Robert RentzschR at
Mon Jul 29 04:02:54 PDT 2013

Hi Oleg & everyone,

3 quick and simple questions here, at least I hope so:

1) in the original Vina paper's benchmark, all RMSDs given (or points drawn) for Vina refer to the top-scoring pose, correct?

2) to be sure: the 'standard error' value given in the text based on the data in Figure 7 does not refer to the most common - I'm not an expert but think so - usage of the term in maths but instead simply to the average (absolute) difference between the predicted:experiment score value pairs, correct?
(btw, is this data available in numbers, e.g. Excel style or st? then I could have checked myself)

3) has it been considered to re-train Vina on a larger dataset, given the increase in structures (e.g. PDBbind refined part in its current state) and do you think it would increase performance? Also, how difficult would it be to retrain Vina in general, on-demand, say one wanted to train on only a particular kind of interaction, such as protein:small peptide. I guess this question is very much related to having alternate scoring functions and things like smina...mmm [sorry this one is the less simple one, as I realise now :)]

Thanks a lot,


Dr Robert Rentzsch
NG4, Bioinformatik
Robert Koch Institut
Nordufer 20
13353 Berlin
Tel: 0049 (0)30 18754 2198

More information about the autodock mailing list