Results and replication Protein–Protein Docking Benchmark 5.0

[lightdock]

Results and replication Protein–Protein Docking Benchmark 5.0

[JorgeRoel]

Hi @knuser,

Thanks for your interest in LightDock. Of course, we do have the results of the paper you mention, but not on a public repository due to storage restrictions (thousand of models * 230 cases). You can just send us an email and we'll find a way to send them to you. In the meantime, you can find the results of the newest 55 entries of BM5 from this publication: https://academic.oup.com/bioinformatics/article/36/3/950/5550626 at https://github.com/lightdock/lightdock_bm5 as well as the scripts to generate all the PDB files. Please, bear in mind that version=0.9.2 includes several improvements compared to version=0.7.0 (the version used in the paper I shared with you), so that the numbers might not be identical between versions. If you want to replicate the results, you would need to install the exact same LightDock version.

With regards to scoring, in general, docking scoring functions hardly correlate with binding affinities, so comparing scores between different systems as you did (A-B vs A-C) by using docking is usually not a good idea. There are other tools (e.g. binding affinity predictors) out there better suited for this task.

And the answer to your last question is no, scores will vary between systems, and again, in general one might expect lower (or higher as in our case) scores for larger interfaces so that, comparing between different systems or setting absolute thresholds might be inappropriate.

Cheers.