KEYWORDS: STRUCTURE_PREDICTION EXPERIMENTAL_DATA

Written by Lei Shi. Ray Wang drafted the previous version.

---

We will use the chemical shifts to improve the fragments from which Rosetta builds up structures, and the NOEs to guide the Rosetta calculations towards the native structure.

Please see references at:

• Rosetta abinitio: Bradley, P et al Science 2005
• Chemical shift fragments: Shen Y et al. PNAS 2008;105:4685-4690
• Chemical shift+NOE+RDC: Raman S, et al Science 2010

In this demo, we will use PDB 2JY7, which is a small protein (for demo purpose) and has experimental data deposited. Several scripts are provided in the scripts folder for formatting purposes:

bmrb2talos.com
cst_map_toCB.py
upl2mini.csh
scores.score.cfg

If you are from David Baker lab, there are scripts available to make setup easier without going through public servers. The following instructions should work just fine without having direct access to any Baker lab cluster.

Running the Demo

1. Create following folders:

   mkdir starting_inputs
   mkdir rosetta_inputs
   mkdir rosetta_inputs/talos_output
   mkdir rosetta_inputs/pick_cs_fragments

2. Download protein fasta and experimental data:
   Download fasta from http://www.pdb.org/pdb/explore/explore.do?structureId=2JY7

   wget http://www.pdb.org/pdb/files/fasta.txt?structureIdList=2JY7 -O starting_inputs/t000_.fasta

   Download chemical shift data from http://www.bmrb.wisc.edu/data_library/summary/index.php?bmrbId=15591

   wget http://rest.bmrb.wisc.edu/bmrb/NMR-STAR2/15591 -O starting_inputs/raw.cs.bmrb

3. Format data for Rosetta use:
   Formmatting chemical shift data for TALOS. The required bmrb2talos script is part of the Rosetta toolbox, downloadable from The CS Rosetta web site. The output file is provided here.

   $ bmrb2talos starting_inputs/raw.cs.bmrb > rosetta_inputs/cs.talos

4. Generating talos predictions using http://spin.niddk.nih.gov/bax/nmrserver/talosn/ using rosetta_inputs/cs.talos. Save/copy pred.tab and predSS.tab to rosetta_inputs/talos_output

5. Generate fragment/profile from RobettaServer http://www.robetta.org/fragmentqueue.jsp using starting_inputs/t000_.fasta. Save/copy t000_.checkpoint to rosetta_inputs/

6. Pick fragments using secondary structure profile and chemical shift data: (where $ROSETTA3=path-to-Rosetta/main/source and $ROSETTA3_DB=path-to-database)

   $> $ROSETTA3/bin/fragment_picker.default.linuxgccrelease -in::file::vall $ROSETTA3_DB/sampling/small.vall.gz -frags::n_frags 200 -frags::frag_sizes 3 9 -frags::sigmoid_cs_A 2 -frags::sigmoid_cs_B 4 -out::file::frag_prefix rosetta_inputs/pick_cs_fragments/frags.score -frags::describe_fragments rosetta_inputs/pick_cs_fragments/frags.fsc.score -frags::scoring::config scripts/scores.score.cfg -in:file:fasta starting_inputs/t000_.fasta -in:file:checkpoint rosetta_inputs/t000_.checkpoint -in:file:talos_cs rosetta_inputs/cs.talos -frags::ss_pred rosetta_inputs/talos_output/predSS.tab talos -in::file::talos_phi_psi rosetta_inputs/talos_output/pred.tab

   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!

7. Run Rosetta with the fragments chemical shift fragments:

   $> $ROSETTA3/bin/AbinitioRelax.default.linuxgccrelease -in:file:fasta starting_inputs/t000_.fasta -file:frag3 rosetta_inputs/pick_cs_fragments/frags.score.200.3mers -file:frag9 rosetta_inputs/pick_cs_fragments/frags.score.200.9mers -nstruct 1 -abinitio::increase_cycles 0.5 -abinitio::relax -score::weights score13_env_hb -abinitio::rg_reweight 0.5 -abinitio::rsd_wt_helix 0.5 -abinitio::rsd_wt_loop 0.5 -disable_co_filter true -out:file:silent csrosetta.out

IMPORTANT

• Change nstruct to generate desired number of models. Larger is better depending on your available computer time, etc.
  
• change -abinitio::increase_cycles 0.5 to 10! (0.5 is chosen for tesing purposes only)

Processing the output

1. Extract the low energy models:

   grep SCORE csrosetta.out | sort –nk2 | head

   The second column contains the energies of the lowest energy 10 models. Select as the cutoff the energy on the last line.

2. This script:

   cull_silent.pl csrosetta.out “score < cutoff”

   will produce csrosetta.select.silent which contains the lowest models below cutoff.

3. Extract pdbs from selected silent file

   $> $ROSETTA3/bin/extract_pdbs.default.linuxgccrelease -in::file::silent csrosetta.select.silent

4. Check convergence by superimposing the ten low energy models in pymol or your favorite molecular graphics

5. Check convergence by clustering the lowest energy models (see clustering demo for instructions)