Variability in docking scores across multiple runs with the same predicted pose

[bnovakborna]

Hello,

I am wondering why I am seeing variability in docking scores across multiple runs. I have 1000 structures that I have docked a single ligand to, saved the first pose and repeated docking to save the first pose again (without changing the seed number). When I look at the scores their correlation coefficient is around 0.8 with big differences for some of the structures (see the pdf attached). The receptor and ligand files that produce the largest difference (around 5kcal/mol), along with docking script can be found here: https://github.com/bnovakborna/variability_question

It is worth noting that the ligand poses that produce the largest score difference are the same (called state000529-ligand-91.pdbqt and state000529-ligand-92.pdbqt following the link above).

Let me know if you need more information from me.

Thank you!
dock1_dock2.pdf

P.S. I updated the github repository to include the seed, so the difference in scores can be reproduced.

[diogomart]

Thank you for the thorough report!

This is a bug, and it's fixed by #81 from @klgunst. I'll merge the PR soon and release v1.2.4.

[bnovakborna]

Thank you for your quick response! Looking forward to it.

[diogomart]

PR #81 is merged, so the current develop branch does not have this bug anymore.

[bnovakborna]

Great! Thank you, I'll check it out.

[diogomart]

Might have been affected by #200