BNRA course 2004

"From Data to Structures"

Program for the practical
"Structure calculation and validation"

Chris Spronk and Aart Nederveen, May 2004

1. General remarks

2. Setup

• The deposited pdb-file
• NMR distance and dihedral angle restraints in XPLOR/CNS format
• The original paper describing the structure in pdf-format

• Calculate and refine a structure ensemble using CNS
• Analyse the structural properties and general quality of the molecule
  
• Compare to the results described in the literature
• Prepare a final presentation of the results
  

3. Software

3.1 Structure calculation

3.2 Structure validation

3.3 Structure visualization

4. Literature

• P. Guentert. Structure calculation of biological macromolecules from NMR data. Quarterly reviews of biophysics 31, 145-237, (1998).
• C.A.E.M. Spronk , S.B. Nabuurs, E. Krieger, G. Vriend & G.W. Vuister. Validation of protein structures derived by NMR spectroscopy. Progress in NMR spectroscopy, in preparation.
• S.B. Nabuurs, C.A.E.M. Spronk , G. Vriend & G.W. Vuister. Concepts and tools for NMR restraint analysis and validation. Concepts in Magnetic resonance , accepted .

5. Practical aspects

5.1 Time schedule

5.2 Directory setup, files and scripts

Note that most scripts are freely available from the RECOORD database

Table 3. Actions, outputfiles and directories

Action	Runcommand	Start in dir:	Output
Generate molecular topology	generate.sh*	project/	project_cns.mtf**
Generate extended starting structure	generate_extended.sh	project/	project_cns_extended.pdb
Run simulated annealing	refineLongSep.sh	/home/course#/	- structures:   str/      project_cns_1.pdb, ..., project_cns_100.pdb - job files, CNS input and output files:   cnsRef/
Do refinement in water	re_h2o.sh	/home/course#/	- structures:   str/wt/      project_cns_w_1.pdb, ..., project _cns_w_25.pdb - job files, CNS input and output files:   cnsWtRef/
Do validation using WHATCHECK and PROCHECK	validPDB.sh	str/wt	- validation results (summary files):   v_project_cns_w/whatcheck      WHATCHECK_project_cns_w.SUM   v_project_cns_w/procheck      PROCHECK_project_cns_w.SUM
Analysis of violations	calcViolOrg.sh	project/	violations/   viol_project_cns_w_0.3
Summary of violation analysis	analysViol.sh	project/	violations/     viol_results

*For an explanation of  the scripts and its options, type the script name without any command line arguments and press enter.
**Directories are indicated in bold; project in italic (as part of a filename) indicates the project ID that has been assigned to you (see Table 1), e.g. project_cns_1.pdb = 1txa_cns_1.pdb for groups1 & 2, 1kdf_cns_1.pdb for groups 3 & 4 etcetera. Only files relevant to the course are specified.



Finally your home directory will be organized as follows:

Main directory	Subdirectory tree
/home/course#/	- project/	- str/	- wt/	- v_project_cns_w/	- procheck/ - whatcheck/
		- cnsRef/ - cnsWtRef/ - violations/
	- ValidationPractical/	- 1ka3/	- procheck/ - whatcheck/
		- 1i1s/	- procheck/ - whatcheck/

5.3 Further guidelines for running the scripts

• Generate the molecular topology (project_cns.mtf file) from the primary sequence or pdb-file.
• Use the script generate.sh.
• The generation of the topology file may take 1-2 minutes. Check the files that are now created.
• If applicable, check the following (look in the topology file and the original .pdb file):
  
• What do you think is the difference between the .pdb file and the .mtf file?
• What is the difference between atom names and atom types in the topology file?
  
• Protonation state of the histidines.
  Check the protonation state of the HIS residues in both the original .pdb file and in the .mtf file. They should be identical.
• Presence or absence of disulfide bridges.
  Check if you can find them in the original .pdb file. They will be mentioned in de header of the .mtf file as well.
• Occurences of cis-peptides.
  To find out if cis-peptides are present in the original .pdb file, you can load the .pdb file in molmol and then choose Calc -> Angles (select the omega angle ca-c-n-ca). If cis-peptides are present they are incorporated in the .mtf file. Just do 'grep CCIS' on your .mtf file and find out if carbon atoms are present with type CCIS. 
• Generate a starting structure in an extended conformation
• Use the script generate_extended.sh. Now an extended structure will be generated. Check this structure with molmol.
• The starting structure will be the input in the simulated annealing protocol, from this the ensemble of structures will be generated
• Now start the structure calculation (simulated annealing part):
• Run the script refineLongSep.sh. Mind: this script should be run from your home directory, one directory lower than the project directory. The argument should be the project name. For example:
  /home/course4> refineLongSep.sh 1txa
  
• This script does the following for you:
  
  • A CNS parameter file is generated, called run.cns. In this file the protocol that we use is specified.
  • Job files are generated for your models and they are processed consecutively on the bug you are working on. This will take a couple of hours, so use another window to check what is going on in the project directory. By default 100 models will now be generated. In general this will be more.
  • Have a look in the directory cnsRef, here you can find the CNS input and output files together with the job files.
  • When a model is finished, it is written in the directory str. Check the header for violations and energy values. For checking the violations and the energies you can use the aliases 'sortener' and 'sortviol' in the str directory.
  • The restraints that are used throughout the calculations are present in the .tbl files. There can be three different files: unambig.tbl (NOE distance restraints), hbonds.tbl and dihedrals.tbl. Which ones do you have? Have a look in those files and figure out how the experimental potentials (see lecture Bonvin) are built up from those numbers. Also calculate NMR data density for your project: the number of restraints per residue.
• The structure calculation will take a couple of hours, in the meantime try to get familiar with the molmol program using the tutorial. Secondly, you can start reading about the structure in the paper. 
• Optional: you can also redo this whole procedure by making a new dir project_var in your home directory and leave out the dihedral restraints, decrease the upper bounds by e.g. 10 % or add some errors in the restraints files. Copy the restraints and the original .pdb file (now project_var.pdb) in the directory project_var  and start again with the script generate.sh
• When the structure calculation is finished (100 models should be present in the directory str), you can go for the next stage: refinement of your models in water. For this we use the script re_h2o.sh, which works similarly as the previous script. This script does the following:
  
  • In the directory str a new directory wt is created in which the 25 best energy structures are copied. These structures will now be refined and will get names like: project_cns_w_1.pdb, project _cns_w_2.pdb, etc.  Use the alias 'sortener' to find out if the lowest energy structures are now water refined by typing this command in the directories str and str/wt.
  • Check the directory cnsWtRef for the CNS input and output files and the job files.
• For the 25 models that are refined in water we will run whatcheck and procheck. Go to the directory str/wt where the refined models are stored. Now the script validPDB.sh will do the work for you (first use it without argument to learn its use, only validate the water refined models with '_w_' in their filenames). This script waits until the waterrefinement stage is finished and the refined models are available, so you can use it immediately after starting the water refinement stage.
• For calculating the RMSD use the secondary structure elements as specified in the paper and use these to superimpose the models using these regions in molmol. If the regions are not specified in the paper, you can find them in the procheck output.
• Now you can use the scripts calcViolOrg.sh and analysViol.sh to check the distance restraints and dihedral angle violations in your waterrefined ensemble of 25 models.
  

5.4 Guidelines for the presentation

Every group has to present some of their results that were obtained during the course. The time for one presentation is 15 minutes, followed by questions and discussion (~5 minutes). Since there is only limited time for the presentation, try to present your results in a compact manner. 

• Introduction of the project
• Description of the protein: e.g.: fold, function, active sites, biological relevance, special features such as cis-peptides, disulphide bridges, etc.
  
• Description of the NMR data: data density, data types
• Results
• Calculated ensemble of structures
• Characteristics and quality (e.g.):
• Secondary structure elements
  
• Discuss value of rms deviations, energies, violations
• Ramachandran plot
• RMSD of ensemble
  
• Other quality indicators
  
• Discussion and conclusion
• Discuss differences between procheck and whatcheck quality indicators
• Identify problematic regions
  
• Comparison with structure as described by the authors: discrepancies and agreement:
• Quality indicators
• Secondary structure
• RMSD

Appendix 1. Introduction to structure validation

• PROCHECK:
• In your home directory go to ValidationPractical/1ka3/procheck/
• To run procheck on this structure type:
  
  procheck ../1ka3_001.pdb 2
  
  Explanation: Structure 1ka3_001.pdb is compared to X-ray structures of 2 Ångstrom resolution

• WHAT IF:
• In your home directory go to ValidationPractical/1ka3/whatcheck/
• To start the program type:
  
  whatif
  
  
• Set a parameter:
  
  setwif 593 100000
  
  Explanation: “setwif” is needed here to make the output of whatif very long (we need this for 1ka3). It sets the internal parameter 593 to 100000, which means that the output can be 100000 lines (not important to know for now, I just mention this for completeness)
  
  
• Enter the WHAT IF checking menu:
  
  check fulchk ../1ka3_001.pdb ../1ka3_001.pdb
  
  Explanation: “check fulchk” is used to enter the structure checking menu in WHAT IF and starts the full checking of a structure (individual checks can also be run from the check menu if desired). This option asks twice for the filename of the structure that needs to be checked, so it is entered twice.
  
  
• Quit the program when the checks are finished. On the first question enter “n”, and subsequently we stop the program with the “fullstop y”:
  
  n
  fullstop y
  
  
• The output is written to “pdbout.txt” in the directory where whatif was run.

1. Start with the summary at the end of the “pdbout.txt” file to get a quick indication of what might be wrong. In the case of 1i1s you will find a rather remarkable value for one of the validation parameters of which we can say with certainty that the forcefield that was used is bad. Find out which parameter this is. Also use the PROCHECK output to verify this. Have a look at the structure in the structure viewer.
   
   
2. In both cases WHAT IF will complain a lot (the checks are very critical), however in the case of 1ka3, there is one property that the program needs particularly many lines for to list all the problems. Find out which one this is and compare the WHAT IF output to the PROCHECK output describing the same property. Have a look at the structure in the structure viewer.  From the websites describing the program output, can you think of reasons for the differences in output? And can you think of what possibly caused this structure to go wrong in such a major way?
   
   
3. Now you have found the major problems in the structures, try to understand the rest of the output of PROCHECK and WHAT IF (if you have time left).
   
   
4. Have a look at the residue specific Ramachandran plots displayed by PROCHECK. Investigate the green areas (=probability distributions: the greener, the more often a residue is found in this phi-psi combination in the reference data base) and look at the differences for different residues. Can you find residues that are more often found in helices than in sheets?