You are here

Using rosetta 3 to make a fragment library and then use it to model loop of a protein structure

5 posts / 0 new
Last post
Using rosetta 3 to make a fragment library and then use it to model loop of a protein structure
#1

Hi,

I want to know how to generate a fragment library for rosetta ... The method given in the documentation is not the one I want as , I want to retreive fragments of 8 or above residues taken from pdb consisting the loop regions of various proteins not on the basis of any resembling sequence.. Also I want to know what's the structure of this file consisting the fragments (means what does the last 4 columns tell)

1bvy F 24 L E -111.045 104.882 178.947 0.947 12.265 16.922 3 0.000 P 1 F 1
1bvy F 25 L E -93.501 124.052 178.069 0.947 12.265 16.922 3 0.000 P 1 F 1
1bvy F 26 V E -112.031 118.672 179.847 0.947 12.265 16.922 3 0.000 P 1 F 1
1bvy F 27 L E -108.327 139.919 178.415 0.947 12.265 16.922 3 0.000 P 1 F 1
1bvy F 28 Y E -130.990 151.362 163.536 0.947 12.265 16.922 3 0.000 P 1 F 1
1bvy F 29 G E -119.351 106.868 -174.261 0.947 12.265 16.922 3 0.000 P 1 F 1
1bvy F 30 S L -136.861 147.527 172.036 0.947 12.265 16.922 3 0.000 P 1 F 1
1bvy F 31 N L -110.802 -36.965 -179.691 0.947 12.265 16.922 3 0.000 P 1 F 1
1bvy F 32 M L -104.530 8.189 168.107 0.947 12.265 16.922 3 0.000 P 1 F 1

What does this 3 0.000 P 1 F 1 stands for ... Coz I am planning to create such a fragment library ..

Post Situation: 
Sun, 2011-01-23 05:22
bharat_46010

Hi,

A) The Rosetta fragment stuff is getting overhauled between 3.1 and 3.2, I think. I've emailed the new fragment guy to see if he has useful comments.

B) Check out src/core/fragment/ConstantLengthFragmentSet.cc, function read_fragment_file. It looks like the columns are:

pdb chain aa_index aa ss phi psi omega

and the rest of the columns are unused. They're probably legacy stuff left over from Rosetta2 that wasn't ported.

C) Rosetta3.1 has an internal fragment picker. It is sort-of documented, there's a README in src/core/fragment. I think 3.2 will have a new and completely different one (which I think will be fully documented.)

Mon, 2011-01-24 07:31
smlewis

Hi,

I am still not clear about creating a fragment library ... I have submitted my sequence in robetta server for creating a fragment library but it has not asked for the size of fragment to be searched for.

Also I was reading the following documentation http://www.rosettacommons.org/guide/Fragments about how to create a fragment library , but there also its no where mentioned about defining the size of fragment to be searched.

So , I am stating my query here again .. I want to create a library of loop fragments of increasing length from 8 onwards .. Can anybody guide me how can I achieve it using rosetta.

Mon, 2011-01-24 22:25
bharat_46010

Hi,

I hope we get the new fragment picker in Rosetta 3.2 and then your problem will be very easy to solve. Yes, you will be able to pick any fragment size you want. The file format you mentioned will become obsolete as probably no one remembers what "0.947 12.265 16.922 3 0.000 P 1 F 1" actually means. But I would guess it is:
- fragment sequence score
- fragment SS score
- total weighted score of a fragment
- P 1 means the first position within a fragment
- F 1 should be the first fragment
But the last two fields look strange in your case (it is only P1 F1, I would expect continuous numbering)

Anyway, you may disregard anything after Phi, Psi, Omega - Rosetta doesn't read it in.

Dominik

Thu, 2011-01-27 00:04
dgront

I've contacted one of the fragment developers, he said he'd post a deeper reply when he had time. I've already told you all I know about it...

Robetta makes fragments of length 3 and 9 (and possibly 1 or 13, I forget). Someone determined those were optimal for ab initio a long time ago and it's never been changed. You may be able to get away with trimming the 9mer library to 8mers.

I'm not aware of any deep documentation for the fragment picker that's in 3.1 (as I said, src/core/fragment). You can play around with that code. What you want to do will certainly require writing C++ code in 3.1. If you wait for 3.2, then it will probably become easier...but it's still fairly likely you'll need to write stuff in C++.

I've attached a bit of code that ought to be mostly-compatible with the 3.1 release. It's from a module that wasn't in 3.1 (and won't be out until 3.3), but it uses the most basic of the fragment-picking commands which ought to get you started. (You can ignore the parts about loops and anchors - those are related to the application, not the fragment picking)

Tue, 2011-01-25 07:42
smlewis