You are here

Unable to understand fragment picking tutorial points.

5 posts / 0 new
Last post
Unable to understand fragment picking tutorial points.

Dear Friends,


I am performing abinitio structure prediciton of a 382 long protein sequence by following the tutorial:

In this tutorial, there is a "fragment picking" tutorial::

Could you please help me understand the below bullet points of the tutorial? I am unabvle to link these below points with what I am supposed to do. Uptil psipred I  understand and have got .ss2 file. I have difficulty in uderstanding:


a) "put this vall in the flag"; where to get the vall from?

b) and, what does "flags" mean here? I think there is something missing in between in the tutorial. 

c) how to get a vall file like mentioned here: filtered.vall.dat.2006-05-05.gz?



After psipred finishes get the "download the machine learning scores results in plain text format" and give it a meaningful name, e.g. 2jsv.psipred.ss2. This should have a header with the string "VFORMAT"

  • Put this vall in the flags:

    $ <path-to-Rosetta>/main/database/sampling/filtered.vall.dat.2006-05-05.gz

    Update the flags with your correct database location

  • Run the picker like this:

$> $ROSETTA3/bin/fragment_picker.linuxgccrelease @BestFragmentsProtocol/flags -in::file::vall $ROSETTA3_DB/sampling/small.vall.gz

IMPORTANT The small.vall.gz used here for fragment picking is only used to speed up the demo. You have to change this to the vall database on your system!


  • frags.200.3mers: for use with -in:file:frag3
  • frags.200.9mers: for use with -in:file:frag9
  • fsc files for fragment information.


Thank you for your time!

Post Situation: 
Thu, 2019-09-05 03:44

Are you interested in 'tweaking' the fragment picker? or just running it?

because running will do everything for you.  the tutorial is just a short part of what does, and is meant for a live demo session where people can run things quickly


please note that the bakerlab fragment picking server located at uses custom sate of the art settings/databases and is always evolving.  That is why most developers choose to let the server do it for them.

comically though -> it appears to be down right now : (

Thu, 2019-09-05 07:58

I figured out what these bullet points mean, but as you mention, now the questions is: 


What if I want to generate fragments from a "Vall" file generated from my specific set of PDBs? The vall file provided by rosetta package is very old "vall.jul19.2011.gz" under the folder ""fragment_tools/". Is it ok to use that vall file for fragment generation?

Even though, if I want to do that and generate my vall file, running is a challenge at the moment for me, because I can't run "make" in "pdb2vall/structure_profile_scripts/DEPTH-CLONE-2.8.7/" to generated "depth" executable to do so; it goves weird erros (have asked a separate question regarding that in this forum ""make" error in "pdb2vall/structure_profile_scripts/DEPTH-CLONE-2.8.7" "). Any idea how this can be solved? 

And, to generate fragment files only, where can I do that? I know robetta serve ( for full abinitio structure prediction but, is there any server just to make fragments?



Thu, 2019-09-05 10:31


Yes that vall is the correct database (under fragment_tools) to use.

Although i do not have  the data myself, i was told that newer iterations of the VAL database caused worse fragments to be generated, but no one could say why : /


As for pdb2vall -- I've never used that before.  I will look into it for you though.

Fri, 2019-09-06 10:24

i responded to your other thread as well but i think we package you already built binaries for DEPTH_CLONE so i think you can skip that step

Fri, 2019-09-06 11:29