Different result/HADDOCK Score each time the same molecules (Protein-DNA) are docked

1

Hi All

I tried docking the same Protein A with the same dsDNA B several times, but I got different results (all the parameters were kept same, HADDOCK Easy Interface). Please help. Aren’t the docking results reproducible?

Thanks and Regards

2

You will only get exactly the same results if you run on exactly the same computer.

HADDOCK distribute jobs to a grid of computers around the world, meaning each job might run on a different hardware.
This is way you will see difference. Those should however not be completely different.

Define what you mean by different results? Are the models looking completely different?

3

Hi

Thanks for your swift reply. I submitted a docking job (exactly same parameters and molecules) one after the other (immediately after). The following is the result that I am getting.

image
image791×235 7.41 KB

Clusters 1.3 and 1.4 have very different energies.

Question 1.The HADDOCK scores suggest to select the cluster with most negative HADDOCK score. But, within each cluster, 4 structures are given. Please suggest which structure to select among them as I have to proceed with one structure for further analysis.

Question 2- Which of the two runs should I take?

Thanks in Advance.

4

Hi there

Submitting immediately after will not change anything. You jobs are still distributed over a wide grid of servers.

You should rather compare the cluster HADDOCK scores on the result pages.

Now you are concentrating on one cluster and its top4 members. The differences in your table are to me not that large considering the noise in those energies.

And I would never put all my money on a single model. Why not analyse all four for example?
Did you actually look at them? Are they that different?

Don’t trust number alone!

And as last, in principle the best prediction would be the top model of the top cluster. But differences with other clusters might not be significant and you will have to consider those as well.

And forget about this binding energy - it is there for historical reasons, but I would not trust (we should actually remove it from the output).

1 Like