Determination of the protein number of states

In this tutorial we will provide you with guided examples for determination of the protein number of states.

In summary, our approach consists of following steps:

1. We conduct a series of structure calculations with 1-9 states
2. We evaluate structure calculations in terms of target function and correlations (using PDBcor)

Software installation

This tutorial requires following software:

1. CYANA
2. Python3
3. UCSF Chimera
4. PDBcor. In case if PDBcor is not yet installed, please go to this link and follow installation instructions. Later we will refer to the PDBcor installation path as PATH/To/PDBcor

Data preparation

Please follow the following steps:

1. Download the demo data.
2. Unpack the demo data

Execution

We recommend to use parallel computation for the execution as it will significantly reduce the total running time.

Please follow the following steps carefully (exact Linux commands are given below; you may copy them to a terminal):

First, edit the protein specific data in the folder data/. Make sure, that filenames are kept the same as in the demo. Advanced users can ignore it and fix PREP.cya for potential naming errors. Change the protein sequence in the root folder (same filename is not required).

Additionally, do following checks:

Check that init.cya and CALC.cya have rmsdrange variable (range to calculate RMSD of the protein bundle) set to correct value

Check that SPLIT.cya has nres variable (number of residues) set to correct value

Then, test the data/ folder by running a test calculation:

cp -r data test
cd test
cyana PREP.cya nbundle=3
cyana -n 20 CALC.cya #cyana CALC.cya if no parallel computing is available

Check that both PREP and CALC scripts run without errors.

Check in program Chimera that resulting structure bundle.pdb is correct.

Check that split structure splitall.pdb is split correct (each conformer 1-nres is a separate PDB model).

Then, edit RUN.cya:

Edit the cyana engine

cyana:=cyana (do it only if you are sure what it is)

If you do not have an access to parallel computation change the line of RUN.cya from

system "pwd; ls -l CALC.cya; $cyana -n 20 CALC inputseed=$seed"

to:

system "pwd; ls -l CALC.cya; $cyana CALC inputseed=$seed"

Then, run the series of two-state structure calculations:

cyana RUN.cya

This creates a folder for each multi-state structure calculation. Final structure is saved as bundle.pdb and final split structure (where each state is an independent model is saved as splitall.pdb).

If you do not have an access to parallel computation change the line of ANALYSE.cya from

system "bsub python PATH/To/PDBcor/correlationExtraction.py $dir/splitall.pdb --nstates=$ns --therm_iter=5"

to:

system "python PATH/To/PDBcor/correlationExtraction.py $dir/splitall.pdb --nstates=$ns --therm_iter=5"

Then, activate the PDBcor environment and run the correlation analysis:

source PATH/To/PDBcor/venv/bin/activate
cyana ANALYSE.cya

This creates a correlations/ subfolder in each structure calculation folder with a correlation value that can be read from the correlations_backbone.txt file.

Since all structure calculations were done for different number of states their target functions are not directly comparable. We need first to normalize them.

For this execute bash script

sh normalize_tf.sh # it usually takes couple of minutes

Once the script is finished run it again and save the output

sh normalize_tf.sh > TF.txt

This file contains normalized target functions

Finally, collect results in a single file:

cyana STAT_COLLECT.cya

This collects all correlation and target function values of all executed calculations in a single file using bash script extract_cor_value.sh into the output file results.txt.