On Wed, Mar 11, 2009 at 6:03 AM, Eldon Ulrich <elu@xxxxxxxxxxxxx> wrote:
Hi,
The attached file contains a draft of two saveframes and their tags for the
RELAX output. I wanted to get this out quickly and will try to send a
version with example values as soon as I can. For the order parameter
saveframe, I hope the relationships between the tags and the values
(descriptions) in the XML output are clear. For the '.model_fit' tag values
like 'S2, te' including the quotes would be appropriate.
I'll get to the tags in a second email, and hopefully cover and check
everything there. For largish documents such as this word file, it
would be appreciated if you could attach this to the relax BMRB 'task'
(https://gna.org/task/?6438) and then the link posted. The reason is
because the attachment is sent out to many people and then also
archived all over the place on the internet (e.g.
https://mail.gna.org/public/relax-devel/2009-03/msg00019.html and
http://www.mail-archive.com/relax-devel@xxxxxxx/msg01845.html) so that
with file attachments, the system gets a bit flooded. Cheers.
For tag values, can these be restricted to fixed values? While it
might be ok for humans to read, it would be extremely useful for all
of the BMRB data to be easily parsable by current or future software.
Having to code a parser that handles any combination allowable for
that tag value would be a nightmare. For example for the same model
you might get:
'S2s, tau_s, S2f, Rex',
'Ss2, ts, S2, rex',
'S2_slow, tau_slow, S2_fast, Rex'
from different programs in the future. Other tags such as
_General_relaxation_list.Relaxation_coherence_type would also really
benefit (well the program parsing the BMRB data would). Could this be
documented in the 'Examples' or 'Help Text' sections in the 'Tag
Details' frame of
http://www.bmrb.wisc.edu/dictionary/3.1html_frame/frame_index.html?
The ADIT-NMR deposition system on the BMRB website can be used to test the
uploading of the RELAX output NMR-STAR files. Just, please do NOT click on
the deposit button. I tried the test file and it was imported into the
system successfully.
Ok, this will be useful. Thanks. I just assumed there was a back end
for testing.
The two earlier e-mails are given below (I am not in the habit of hitting
'reply all'), but I think going with the information in the attached
document may be the best at this point, but I am open to suggestions.
That's ok, I'm constantly reminding everyone and need reminders myself
from time to time.
Eldon
--------------------------------------------------------
Hi,
You are correct that the original BMRB NMR-STAR was developed from reviews
of published literature and discussions with Art Palmer many years ago. The
world certainly changes. I wonder if it is worth redoing the NMR-STAR model
almost from scratch? I would prefer taking this path myself, but will need
considerable help from you, if possible. If something new is generated, I
would like it to be general and not tied to on particular software package.
What do you think?
It should be theory dependent rather than software dependent and
flexible enough to handle future developments. I'll get to that in
the next post, because there could be a clash here with SRLS theory
and other future theories.
Eldon
Hi Edward,
One method we have used in the past to model data that is highly related,
but may have many different sources (derived using different software
packages with different configuration parameters) is to create a table where
one of the columns is the name of the value. This is far more flexible than
the current situation with order parameters, where a specific tag is used.
An brief example in NMR-STAR speak might look something like this:
loop_
_Order_parameter_variable.name
_Order_parameter_variable.value
'Tau e' 10-6s
'Tau e err' 10-7s
stop_
The tag '_Order_parameter_variable.name' can be enumerated with a large
number of values and the enumeration can be left open so that new terms can
be accepted at the time of deposition and incorporated into the 'official'
enumeration list later.
There are only 2 equations, the original Lipari-Szabo 1982 model-free
{S2, te} and the Clore et al. 1990 extended model-free {S2f, tf, S2s,
ts} equations. Add Rex, and maybe in the future CSA and bond length r
into the parameter mix, and that is the totality of all parameters
(maybe also the Rex alpha value as well). So for model-free, the
current fixed tags should be sufficient and then maybe new tags added
for new parameters as people invent them?
Oh, actually one model is the sum of all model-free models for all
spins together with the diffusion tensor. Maybe this could go into
the diffusion tensor, or maybe it's not needed at all? I might need
to think this over when I try to work out how to dump the analysis
procedure used (which can be extremely diverse).
Cheers,
Edward
Re: To all relax developers: presenting relax at the ENC conference and preliminary BMRB NMR-STAR file creation for model-free results.