Rosetta-Holes Scoring Function

[Anonymous]

I have been trying to use the scoring function to calculate the rosetta-holes score for a pdb file (Initially I am using the pdb provided in the scoring demo). I keep getting an error related to reading the input file. I think this is the input that is passed to DAlphaBall, but it might be some other problem I am unaware of. I am using the following command:

score_jd2.linuxgccrelease -s test_in.pdb -score:weights holesscore -database ../../../rosetta_database/ -holes:dalphaball "pathToDAlphaBallExe"

and I get the following error (repeated for 10 tries):

core.scoring.packing.compute_holes_score: compute_holes_surfs try: 1
Problem reading input file
core.scoring.packing.compute_holes_score: DAlphaBall output index mismatch 13458528 1 1 140651699073829
29

Any help on what I am doing wrong would be much appreciated. Thanks.

[Anonymous]

I’ve forwarded this to the Holes developer.

[Anonymous]

what is the file holesscore?
is the “pathToDAlphaBallExe” a .o (c object file)?
thanks in advance

[Anonymous]

Will Sheffler (the author) says:

what you will need to do is build the following fortran code.. there
is a crude makefile in the src directory, along with linux and mac
binaries that *might* work. you then need to pass rosetta the
following option: -holes:dalphaball to use
rosettaholes in calculations.

http://dl.dropbox.com/u/3480832/DAlphaBall_new.tar.bz2

I believe that the “holes” app isn’t in the public release, so you
will also need the attached two files, which should be put here
(replacing the first):
rosetta-3.3/rosetta_source/src/apps.src.settings
rosetta-3.3/rosetta_source/src/apps/public/holes.cc

something on command line options:
http://dl.dropbox.com/u/3480832/How_to_Run_RosettaHoles_4up.pdf

[attesor]

I use rosetta 3.3 and 3.4. After some effort, I got DAlaphBall work on my computer. But trying

score_jd2.linuxgccrelease -s test_in.pdb -score:weights holesscore -database rosettaDB -holes:dalphaball MyDAlphaBallExecutable

yields:
“ERROR: Unable to open weights/patch file. None of (./)holesscore or (./)holesscore.wts or /Path2Rosetta/rosetta3.4/rosetta_database/scoring/weights/holesscore or /Path2Rosetta/rosetta3.4/rosetta_database/scoring/weights/holesscore.wts exist”

I tried to search scorescore in the rosetta directory but failed to find anything. Should I download the weight files from somewhere?

[Anonymous]

Hi all,
Those attachments mentioned by Will Sheffler are still ot available.
Would be nice if they were.
cheers
Dave

[Anonymous]

RosettaHoles requires an external application, DAlphaBall (Edelsbrunner E, Koehl P, Combinatorial and computational geometry, Vol. 52. Cpp 243–275.), to do surface/volume and derivative calculation. The “pathToDAlphaBallExe” is just that – the filesystem path to the (compiled and runnable) DAlphaBall executable.

[Anonymous]

Hi

I tried to include holes 1.0 in my .wts file and got DAlphaBall going but I’m not able to reproduce numbers from the RosettaHoles2 article. Also, the holes score reported in the pdb file produced by fixbb (with holes in wts) does not respond to -holes:minimize [decoy25|resl]. I’m not sure what score I’m looking at. For example, for 1ubq I get 4.77735.

Would it help to get the holes.cc app mentioned here and is this possible?

Thanks

[Anonymous]

Am I missing something? How do we get these “attachments”? Has anything changed for running this with 3.4?

[Anonymous]

The attachments got eaten in the system update…the admin is aware but I don’t know if we’ll get them back or not. I haven’t coordinated with Will yet on Holes, in reply to your other thread – he asked me about the release schedule but didn’t say he had documentation?

[Anonymous]

There really isn’t a canonical “holesscore.wts” weights file for scoring with holes, even in the unreleased development version – at least not that I can find.

There’s actually several versions of the holes energy function in Rosetta. There’s “holes_resl”, which was trained on resolution data, “holes_decoy”, which was trained on decoy discrimination, and “holes_min”, which was also trained on decoy discrimination, but at a finer-grained level. There’s also derived terms “holes_min_mean”, which is holes_min normalized by the number of balls, and plain “holes” which is a non-linear combination of holes_resl and holes_decoy.

You should probably start with holes_min, though another of the terms might work better for your particular situation. The holes scores are typically rather large numbers, so a small weight (0.01 or so) probably would be a good starting point. Just add a line turning it on to a copy of your favorite scorefunction, like rosetta_database/scoring/weights/score12_full.wts

[Anonymous]

I believe this is the “holes.cc” file mentioned. (Remember to change the extension when you save it to src/apps/public/holes.cc

For the apps.src.settings file, your best course of action is to edit the existing file, and add a

"holes",

line to the file in the “public”: block. — Between “fragment_picker” and “idealize_jd2” would be a good place for it.
Then just recompile and you should have the holes executable.

[Author]

Posts