Hello,
Tim just reminded me he had some comment on this which I forgot to forward
to the mailing list:
The system assumes that peaks are picked and linked to the amides, and
that once this is done the i/i-1 setup is done via the [Find Resonances
From Peaks] button in the Spin Systems tab of the automated Seq. Assign
GUI.
Also, the system is basically not designed for web services, for the user
to run PSIPRED separately or to make use of BLAST on a sequence database.
If these things are deemed to be essential, for whatever reason, then
people will just have to wait until the WMS system comes online.
Rasmus may be persuaded to add some extras, but this was more work than I
was willing to do at the time. And in any case some files may well be
overwritten by the system if the user inserts their own files. I think
that gaps in the sequence or missing peaks should work fine in general,
but we can only guarantee such things if the CCPN code is running
everything. One cannot bypass our system and expect things to be easy.
Wayne
On Fri, 9 Dec 2011, Martin Ballaschk wrote:
> So no answer means:
>
> a) no one uses MARS
> b) that was way to much text in that help request
> c) everybody who knows decided not to answer because overlong requests need to be punished ;)
>
> So, I'd still be delighted if anyone could share her experience.
>
> Regards
> Martin Ballaschk
>
> On 02.12.2011, at 14:47, Martin Ballaschk wrote:
>
>> Hi there.
>>
>> Using the MARS wrapper of CCPN Analysis 2.2 remains so far an unsolved challenge to me.
>>
>> Problem Nr. 1:
>>
>> CB and CA resonances need to be typed before fed into MARS, and MARS obviously needs to know if the resonance belongs to the current (C, CB, CA) or the preceding (C-1, CB-1, CA-1 etc.) residue. I don't see a way to assign that within CCPN. All I can do is typing the resonance inside the assignemnt popup, by making a new carbon resonance, setting it to the "Same Spin System" and give the type C, CA, CB, etc.
>>
>> Also from a fully assigned protein, where the sequential information is definitely there, the preceding residue shifts seems *not* to be easily exported from CCPN into MARS. The corresponding "-1" columns are empty in the exported CS tables, which can be found in the "mars" directory inside the project directory.
>>
>> What am I doing wrong here?
>>
>> Problem Nr. 2
>>
>> For some reason, CCPN demands a local install of Psipred to generate the secondary structure information which also can be obtained via the Psipred web service. CCPN also demands the .ss and .ss2 files that a local installation of Psipred is generating, but MARS does not use anyway. (It just needs the regular ".horiz" file.)
>>
>> %%%
>>
>> I've run in to several other problems, e.g. will blastpgp – which is used by standard Psipred fail silently, but the experimental Psipredplus using BLAST+ works fine. Also, my Mac and its tcsh has some problems with the standard Psipred scripts.
>>
>> Another problem is that the Mac OS MARS binary provided by Markus Zweckstetter is compiled with support for PowerPC processors, which are not built into any Apple computers for more than five years now, IIRC. In the last version of the Mac Operating System, PPC support was dropped and will not return.
>>
>> Below you will find a list of what I've tried so far.
>>
>> I hope someone can help me with this. How are you doing it?
>>
>> All the best
>> Martin Ballaschk
>>
>>
>>
>>
>> %%%
>>
>> Howto (not) use MARS automated protein sequence assignment with CCPN Analysis
>> =============================================================================
>>
>>
>> 1.) Setup Psipred and BLAST+
>>
>> Download and install NCBI/BLAST+ binaries
>> ftp://ftp.ncbi.nlm.nih.gov/blast/executables/LATEST/
>>
>> 2.) Setup experimental version PSIPRED running with BLAST+
>>
>> Download PSIPRED from here:
>> http://bioinf.cs.ucl.ac.uk/software_downloads/
>>
>> Make & Install PSIPRED according to Readme
>>
>> Download uniref90 in FASTA format (4 GB)
>>
>> prepare uniref database according to README:
>>
>> $ pfilt uniref90.fasta > uniref90filt
>> $ formatdb -t uniref90filt -i uniref90filt
>>
>> Set all the paths in the psipredplus script and run it on your protein sequence (FASTA format):
>>
>> $ ./pipredplus myprotein.fasta
>>
>> The resulting files are ...
>>
>> myprotein.horiz
>> myprotein.ss
>> myprotein.ss2
>>
>> ... which can be placed inside the psipred directory of the CCPN project, in case the Psipred wrapper of CCPN does not work.
>>
>> set the following variable in ~/.cshrc
>> setenv PSIPRED_DIR /path/to/psipred_dir
>>
>> 3.) Setup MARS
>>
>> Get the MARS binary (Mac PPC, Linux) from Markus Zweckstetter's Website (MPI BPC)
>> http://www.mpibpc.mpg.de/groups/zweckstetter/_links/software_mars.htm
>>
>> set PATH variables in ~/.cshrc
>> setenv MARSHOME /path/to/mars/bin
>> alias runmars $MARSHOME/runmars
>>
>> Try a expample data set from examples directory to check if all paths are set correctly.
>>
>>
>> 5.) Get CCPN 2.2 (today a buggy "leading edge release")
>>
>> Assignment > Automated Seq. Assignment > Automation
>>
>> If all environment variables are set right, you should have the choice between Nexus (built-in) and MARS. Set all the parameters according to your spectra and protein properties.
>>
>> 6.) Hit "Run Mars"
>>
>> CCPN will ask Psipred to make all three secondary structure files generated by the local Psipred; however if you place them into the "psipred" directory inside the project directory, it will accept them.
>> This is done by a little script sitting in CCPNPATH/ccpnmr2.2/ccpnmr2.2/python/ccpnmr/analysis/wrappers/Psipred.py
>>
>> Next step is CCPN exporting a MARS compatible chemical shift list. This is done by the built-in export module.
>>
>> Next is MARS run by a script CCPNPATH/ccpnmr2.2/ccpnmr2.2/python/ccpnmr/analysis/wrappers/Mars.py
>>
>> 7.) Watch the program fail.
>>
>> The number of chemical shifts has to be identical to the number of amino acids in my protein - what about peaks disappearing due to conformational or proton exchange?
>
|