- modeller_usage - salilab.org

SEQUENCE_COMPARISON: wrong calculation for variability
by Timothy A. Springer Ph.D. 14 Feb '99

14 Feb '99

I like the sequence comparison functions in Modeller. I would like to calculate sequence conservation or variabilty as a function of position. One of the proteins I work with contains a conserved domain, and I would like to plot this. Unfortunately, variability as a function of alignment position is not calculated in a meaningful way. It does calculate what it says it calculates, but the result is not helpful at all - it is the RMS deviation of all residue - residue scores. I have plotted this, and it shows little relationship to highly conserved areas. It picks up positions of identity fine, but that is about it. Could you please implement a calculation of sequence conservation. I suggest the total of the similarity scores. You could divide this by the maximal number of sequence comparisons. I would not divide by the number of comparisons at each position, because this may be small for positions with gaps, and might overestimate conservation in gapped regions. Also, the documentation says gaps are included in the scoring matrix, but this is only true for the blosum matrix. Thanks, Tim Springer >From the manual: SEQUENCE_COMPARISON -- compare sequences in alignment Description: The family variability as a function of alignment position is calculated as the RMS deviation of all residue - residue scores at a given position, but only for those alignment positions that have at most MAX_GAPS_MATCH gaps. The variability is written to file VARIABILITY_FILE , as is the number of pairwise comparisons contributing to each positional variability. ******************************************************************************* * Timothy A. Springer, Ph.D. * * Latham Family Professor of Pathology * * Harvard Medical School e-mail: springer(a)sprsgi.med.harvard.edu * * Center for Blood Research phone: 617-278-3200 * * 200 Longwood Ave. Room 251 fax: 617-278-3232 * * Boston MA 02115 * *******************************************************************************

1 0

Re: Refinement from my own initial Structure
by Tony Pemberton 12 Feb '99

12 Feb '99

On 28th Jan, I wrote:- Dear Modeller users, I have been having problems carrying out a refinement from my own initial structure. I have adopted the steering file example from FAQ no. 12, but my procedure is resulting in the .ini file being overwritten. I want to produce a model based on some templates, but want to carry out the procedure in two stages. The first stage should be to achieve the overall fold based on the templates. The second stage is to refine part of the target based on another alignment which should produce (I hope) better results in that part of the target. To put it another way, this is a bit like the chimera example in FAQ no. 4, except that I think I can get the overall fold by the two stage process outlined above (the problem in FAQ no. 4). Here is my steering file for the second stage: # Homology modelling by the MODELLER TOP routine 'model'. INCLUDE # Include the predefined TOP routines SET ALNFILE = 'seq.pir' # alignment filename SET KNOWNS = '1abc' '1cde' '1fgh' # templates SET SEQUENCE = '1ijk' # target SET ATOM_FILES_DIRECTORY = './' # directories SET STARTING_MODEL= 1 # index of the first model SET ENDING_MODEL = 3 # index of the last model SET DEVIATION = 4.0 # has to be >0 if more than 1 model SET RAND_SEED = -12312 # to have different models from # another top SET MD_LEVEL = 'refine3' # what kind of optimization is done # Specifiying an alternate starting structure. # Specify the initial structure filename, and tell the program to # read the initial file, not construct it from the templates: # ref.ini is my model from stage one # SET MODEL = 'ref.ini', BUILD_METHOD = 'read_xyz' # Homology modelling by the MODELLER TOP routine 'model'. # Demonstrates how to refine only a part of the model. SET DYNAMIC_FLAG = 0 # selected atoms do not feel the #neighbourhood CALL ROUTINE = 'model' # do modelling on supplied modelling # (hopefully) # select the residues for second stage refinement SUBROUTINE ROUTINE = 'select_atoms' PICK_ATOMS SELECTION_SEGMENT='17:' '43:', SELECTION_SEARCH='segment', ; PICK_ATOMS_SET=1, RES_TYPES='all', ATOM_TYPES='all', ; SELECTION_FROM='all', SELECTION_STATUS='initialize', SELECTION_STEP=1 RETURN END_SUBROUTINE ####################################################### I assumed that "BUILD_METHOD = 'read_xyz'" was a flag to modeller that tells it *not* to construct an initial model from the templates, but to read the .ini file supplied. Correct? (y/n) The result of the above is to construct both a new .ini file with part of the molecule built from topology (extended chain) and the same for the resultant structure, but with the rest refined as I wanted. That is, I get the refinement I wanted but lose the overall fold from stage one. Can anyone tell me where I am going wrong. THE SOLUTION ------------ After reading through the modeller steering files and constructing a flow chart of the procedure, I realised that I was essentially correct about the BUILD_METHOD being a "flag". This acts as a variable but it is the *wrong* variable. It should be GENERATE_METHOD not BUILD_METHOD Therefore the line in the above script should read: SET MODEL = 'ref.old', GENERATE_METHOD = 'read_xyz' This variable is evaluted in __homcsr.top in the 'homcsr' subroutine. The default is 'transfer_xyz' Regards, Tony Pemberton ********************************************************************* Mr. A.J.Pemberton Tel: +44 121-414-3388 c/o Dept. Rheumatology, Fax: +44 121-414-3982 Medical School, E-mail: A.J.Pemberton(a)bham.ac.uk The University of Birmingham, Birmingham B15 2TT. U.K. *********************************************************************

1 0

sequence_to_ali command
by modelisation 09 Feb '99

09 Feb '99

Dears, we have an unexplained problem when using the SEQUENCE_TO_ALI command: only the first residue of the MODEL (template) file is transfered to the alignment array, e.g if we write an alignment file, only the first residue appears, though the right number of residues is written by the program (the second line of the alignment file includes the first and last residue number of my pdb file). Here is an example of the output i get when try to test this command on 5fd1: >P1;all structureX:all: 1 : : 106 : :undefined:undefined:-1.00:-1.00 A* I would like to precise that we get this problem when using the example loop routine, as well as when simply testing the command e.g: READ_MODEL FILE = '5fd1.ent' SEQUENCE_TO_ALI WRITE_ALIGNMENT FILE='5fd1.ali' What did we do wrong? Regards. Michel Giorgi Lab. Cristallochimie Fac. St. Jerome, Marseille, France

2 1

Copper parametrization
by Michel Giorgi 05 Feb '99

05 Feb '99

Hi! After some difficult startings with the MODELLER package i feel now really enthusiastic about the possibilities of the program. Anyway i have a question: did one of you parametrized the copper atom? i.e do you know the radii values we have to enter in the radixx.lib files and the nonbonded constants of the parm.lib file? (what is the meening of these last values?) Maybe as a first (good?) approximation i could use the values of the ZN atom which is parametrized. Anyone has an idea? Regards, -- Michel GIORGI Lab. de Cristallochimie Faculte de St. Jerome Marseille, cedex 20, FRANCE

2 1

[Fwd: Enormous modeller for Linux]
by Andrej Sali 02 Feb '99

02 Feb '99

-- Andrej Sali, Assistant Professor The Rockefeller University, 1230 York Avenue, New York, NY 10021-6399 tel +1 212 327 7550; lab +1 212 327 7206 ; fax +1 212 327 7540 e-mail sali(a)rockvax.rockefeller.edu; http://salilab.org

1 0

Enormous modeller for Linux
by Graham Tyrrell 02 Feb '99

02 Feb '99

I am currently using the ENORMOUS (2000 residue) version of Modeller4 running on an SGI. However, we have recently built a large Linux cluster, and I would like to try running Modeller on it. It appears that the compiled versions for Linux at the ftp site only allow up to 1000 residues which is of little/no use to me as I am modelling a 1200 residue protein. Any possibility of an ENORMOUS version for Linux? (presumably, available memory is a large constraint... what kind of requirements does Modeller have?) Regards Graham Tyrrell -- ---------------------------------------------------------------------- Graham Tyrrell Email (Work) : tyrrell(a)yorvic.york.ac.uk (Personal) : gpt103(a)york.ac.uk Telephone : (01904) 432550 -----------------------------------------------------------------------

1 0

Defining Custom residues
by David Chalmers 01 Feb '99

01 Feb '99

Hi All, I am having some difficulties defining my own residues for use with Modeller4. Despite reading the FAQ and the other documentation Modeller does not seem to recognise my residues. Does anybody have an example of a protein containing a custom residue or ligand that I could study to see if I am doing the right things? Thanks for any help that anyone can provide, David _____________________________________________________________________________ David Chalmers Lab: 9903 9110 Victorian College of Pharmacy Fax: 9903 9582 381 Royal Pde, Parkville, Vic 3053 http://synapse.vcp.monash.edu.au Australia David.Chalmers(a)vcp.monash.edu.au _____________________________________________________________________________

1 0

Refinement from my own initial Structure
by Tony Pemberton 29 Jan '99

29 Jan '99

Dear Modeller users, I have been having problems carrying out a refinement from my own initial structure. I have adopted the steering file example from FAQ no. 12, but my procedure is resulting in the .ini file being overwritten. I want to produce a model based on some templates, but want to carry out the procedure in two stages. The first stage should be to achieve the overall fold based on the templates. The second stage is to refine part of the target based on another alignment which should produce (I hope) better results in that part of the target. To put it another way, this is a bit like the chimera example in FAQ no. 4, except that I think I can get the overall fold by the two stage process outlined above (the problem in FAQ no. 4). Here is my steering file for the second stage: # Homology modelling by the MODELLER TOP routine 'model'. INCLUDE # Include the predefined TOP routines SET ALNFILE = 'seq.pir' # alignment filename SET KNOWNS = '1abc' '1cde' '1fgh' # templates SET SEQUENCE = '1ijk' # target SET ATOM_FILES_DIRECTORY = './' # directories SET STARTING_MODEL= 1 # index of the first model SET ENDING_MODEL = 3 # index of the last model SET DEVIATION = 4.0 # has to be >0 if more than 1 model SET RAND_SEED = -12312 # to have different models from # another top SET MD_LEVEL = 'refine3' # what kind of optimization is done # Specifiying an alternate starting structure. # Specify the initial structure filename, and tell the program to # read the initial file, not construct it from the templates: # ref.ini is my model from stage one # SET MODEL = 'ref.ini', BUILD_METHOD = 'read_xyz' # Homology modelling by the MODELLER TOP routine 'model'. # Demonstrates how to refine only a part of the model. SET DYNAMIC_FLAG = 0 # selected atoms do not feel the #neighbourhood CALL ROUTINE = 'model' # do modelling on supplied modelling # (hopefully) # select the residues for second stage refinement SUBROUTINE ROUTINE = 'select_atoms' PICK_ATOMS SELECTION_SEGMENT='17:' '43:', SELECTION_SEARCH='segment', ; PICK_ATOMS_SET=1, RES_TYPES='all', ATOM_TYPES='all', ; SELECTION_FROM='all', SELECTION_STATUS='initialize', SELECTION_STEP=1 RETURN END_SUBROUTINE ####################################################### I assumed that "BUILD_METHOD = 'read_xyz'" was a flag to modeller that tells it *not* to construct an initial model from the templates, but to read the .ini file supplied. Correct? (y/n) The result of the above is to construct both a new .ini file with part of the molecule built from topology (extended chain) and the same for the resultant structure, but with the rest refined as I wanted. That is, I get the refinement I wanted but lose the overall fold from stage one. Can anyone tell me where I am going wrong, Regards, Tony Pemberton ********************************************************************* Mr. A.J.Pemberton Tel: +44 121-414-3388 c/o Dept. Rheumatology, Fax: +44 121-414-3982 Medical School, E-mail: A.J.Pemberton(a)bham.ac.uk The University of Birmingham, Birmingham B15 2TT. U.K. *********************************************************************

2 1

MODELLER and external constraints (such as NMR)
by chunhua yuan 26 Jan '99

26 Jan '99

Dear MODELLER users: I wonder whether anyone have tried to incoporate external constraints such as those from NMR in modeling. I have some NMR constraints derived from NOE and want to incorporate them into molecular dynamics step, but have no idea how to get start, such as the format (page 72 manual) and the command in TOP file. By the way, I have got the model using "fully automated comparative modeling." Thanks for any comments. Best regards, Chunhua Yuan

2 3

Re: Modeller: too many parameters?
by Andrej Sali 08 Jan '99

08 Jan '99

Also, if you are using more than 1 templae, it could be that they are aligned such that equivalent distances in the templates are more than say 5 angstroms different. Check the output of CHECK_ALIGNMENT in the log file please. If this is so, that would also increase the number of parameters needed for the splined restraints (fixed dx). Andrej szia(a)hanga.enzim.hu wrote: > > Hi, > > > Please try > > > > SET SPLINE_ON_SITE = off > > > > before CALL ROUTINE = 'model'. > > Thanks. The program seems to be running now, apart from the fact that it > writes some hundreds of thousands of warnings of the following type to the > log file: > > chkder___W> DVY(i) out of bounds: 6646 6646 0.1013E+13 > > But Andras Fiser was here the day before yesterday and he told me that > this is not an actual problem. > > > Can I put this onto the MODELLER usage mailing list? > > Yes, of course. > > Best, > Andras > > Andras Szilagyi (szia(a)enzim.hu) > Institute of Enzymology, Hungarian Academy of Sciences > Karolina ut 29., H-1113 Budapest, HUNGARY > Fax: +36 1 166 5465, phone: +36 1 166 5633 -- Andrej Sali, Assistant Professor The Rockefeller University, 1230 York Avenue, New York, NY 10021-6399 tel +1 212 327 7550; lab +1 212 327 7206 ; fax +1 212 327 7540 e-mail sali(a)rockvax.rockefeller.edu; http://salilab.org

2 2