You are here

loop remodelling with a fixed bound DNA

9 posts / 0 new
Last post
loop remodelling with a fixed bound DNA
#1

Hi all,

I am trying to use the loop remodelling app (loopmodel) to refine my homology model of a protein that binds to DNA.

I assume it must be possible at some level to remodel loops with (fixed) non-protein atoms but it seems to be the case that loopmodel only understand protein atoms.

Is there another way of doing this - perhaps I need to use PyRosetta?

thanks very much!
James.

Thu, 2010-02-25 03:23
jtmacd

I would guess that the loop modeling app is throwing out your non-protein atoms. It has some poorly thought out assumptions...

PyRosetta is likely to be an reasonable way around this problem. You can pick apart the loop modeling executeable to see the Movers it calls; if you set up your Pose yourself (and FoldTree, and MoveMap) to handle the DNA and then call those Movers yourself I bet it will work. Worst comes to worst, directly call the KinematicMover and/or (Small/Shear/Frag + CCDLoopClosureMover) directly.

I haven't played around with PyRosetta but I'm reasonably certain all the necessary parts are present.

Thu, 2010-02-25 07:37
smlewis

having now had a look at PyRosetta I'm still not sure this is possible.

If I want to use fragments and CCD my understanding is that you need to convert to a centroid model and then back to full atom for High-Res refinement. Unfortunately I don't think centroid mode understands DNA residue types (at least when I have tried).

So I am able to load my complex and run high-res refinement on my model successfully (I think) but I would really like to do some more aggressive frag set moves beforehand.

Is there any way around this?

P.S. I should point out that my complex is quite large so I have not idealized my models.

Mon, 2010-03-08 09:51
jtmacd

There's no reason I'm aware of that centroid mode is necessary for fragment insertion + CCD. It tends to work a lot better, because spurious/relaxable sidechain clashes are reduced, but it's not necessary.

Go ahead and try fragment moves + ccd + energycutrotamertrialsmover in full atom mode. It ought to work. You NEED that rotamer trials, though, to fix sidechain clashes. (full packing is likely to be too expensive).

Idealization shouldn't be necessary here - it used to be more important, but now the AtomTree stores real lengths/angles instead of looking them up. Idealization is only necessary for the most stripped-down silent files and for some abinitio things, I think.

Mon, 2010-03-08 12:43
smlewis

hmm, sorry to be back again so soon but now I think i've discovered another problem with this approach.

Unfortunately my complex is more than 999 residues long and as far as I can work out this makes it incompatible with the fragment file format
assuming that in the line:

position: 127 neighbors: 200

the "position:" here is the residue index

and that it is specified also here:

3bam A 201 R L -93.490 -14.266 -179.524 0.018 4.874 22.272 4 0.000 P127 F 8

where it says "P127"

But even after I have set up fragments that I think correctly correspond to parts of the complex I would like to remodel I still get the error:

protocols.abinitio.FragmentMover: couldn't find fragment to insert!!

so I think the easiest thing I can do is model each chain separately and then put them back together at the end.

Tue, 2010-03-09 06:47
jtmacd

I agree that this is likely to be a limitation of the fragment reader.

It might be possible to tweak the reader itself to support another digit's worth of residue numbering.

There is a different system available in 3.1, but it's not really documented and it's already been deprecated and replaced in the developer branch. This system lets you pick fragments directly from the Vall (basically a giant fragment database distilled from the PDB). I can give you some code fragments that *might* get you started but I can't support it more than that...

Tue, 2010-03-09 07:46
smlewis

Hi there,

Sure I wouldn't mind trying it out. Is it C++ or Python? I don't mind trying either.

Do you need my email address?

thanks for taking the time to answer my questions.

Tue, 2010-03-09 09:37
jtmacd

C++, and yes, please. I'm not unwilling to paste code in here but the formatting will shatter...

Wed, 2010-03-10 08:42
smlewis

my email address is:

j.macdonaldNO -at- SPAMimperial.ac.uk

obviously removing NO SPAM and putting in the @

thanks very much.

Thu, 2010-03-11 04:47
jtmacd