docking search: first residue centroids, then full atom?

[davidbaker]

I am a novice, learning RosettaDock by trial, error, and reading the web.
My goal is to dock a yeast protein to a kinase — we have other experimental evidence
that the protein in question is a direct substrate of that kinase. I hope that docking can
either contradict, or lend support, to that claim.

My question: how do I locate the ‘funnel’?

My naive approach has been to
1) run hundreds of simple fast searches, with side chains expressed only
as residue centroids
2) find some low-scoring results
3) run those again in full-atom mode (are the full-atom scores still good?)
4) eventually, use the lowest scoring full-atom models for docking in
perturbation mode, hoping to find a funnel

Does this make sense? Or does ‘residue centroid’ docking mode leave out
too much detail to be useful? Finally, is there a guide for all of this? I have
not found one.

Many thanks,

– Paul Shannon
Institute for Systems Biology

[Anonymous]

See the following answers from Docking developers.
—-

Dear Paul,

If I understand correctly, you would like to dock two proteins, one of which is a yeast protein and the other is a kinase. I would go about it as follows:

1. If some biological information is known like active sites etc which are known to be at the interacting interface, align the two molecules in a consistent manner. (Constraint files might be used to bias the simulations) (only -dock_pert 3 8 8 –spin flags are sufficient)

2. If you have no idea about the interacting regions, it does not matter how the partners are aligned ( -randomize1 –randomize2 –dock_pert 5 10 10 –spin)

3. If it’s only a dock_pert around 1000 decoys are sufficient whereas for a randomize run around 100k is a good number ( you might want a smart score filter)

4. At the end look at the lowest scoring decoys. They are probably complex candidates. If on clustering the 200 lowest scoring decoys, a cluster of significant size with the representative member having one of the top (lowest) scores (subjective) is found, that is probably the best guess.

A single low energy (good) point is not sufficient enough to be called a hit and might be a false positive.

Best,

Aroop

---

> I am a novice, learning RosettaDock by trial, error, and reading the web.
> My goal is to dock a yeast protein to a kinase — we have other experimental evidence
> that the protein in question is a direct substrate of that kinase. I hope that docking can
> either contradict, or lend support, to that claim.
>
> My question: how do I locate the ‘funnel’?
>
> My naive approach has been to
> 1) run hundreds of simple fast searches, with side chains expressed only
> as residue centroids
> 2) find some low-scoring results
> 3) run those again in full-atom mode (are the full-atom scores still good?)
> 4) eventually, use the lowest scoring full-atom models for docking in
> perturbation mode, hoping to find a funnel
>
> Does this make sense? Or does ‘residue centroid’ docking mode leave out
> too much detail to be useful? Finally, is there a guide for all of this? I have
> not found one.
>
> Many thanks,
>
> – Paul Shannon
> Institute for Systems Biology

[Author]

Posts