Hi,
Please see my comments below. It's a pity that the code was developed
using the 1.2 line code as there will need to be a migration to the
1.3 line. Note that the branch I created
(https://mail.gna.org/public/relax-commits/2008-12/msg00146.html) is
based on the 1.3 line. This is because the 1.2 line is stable, i.e.
only bug fixes are allowed, and that code is now more than 2 years old
(the 1.2 line started on 2006-01-14 and the 1.3 line on 2006-07-06).
On Fri, Dec 12, 2008 at 2:46 PM, Pavel Kaderavek <kada@xxxxxxxxxxxxx> wrote:
Hi,
we have embodied into the Relax program possibility to perform the
analysis with consideration of the asymmetric character of the CST and
also all dipole-dipole interactions arising from the spins in the
proximity to the studied S-I spin pair.
This should be taken into the account for the dynamic studies of the
fully labeled nucleic acid (Poster 118 at 23rd ICMBRS meeting in San
Diego) and we would like to add this feature to the public version of
the Relax program.
As long as the relax conventions are followed, as described in the
development part of the manual, the code does not break the other
parts of relax, and there are no copyright issues, then any code for
dynamics will be considered for relax. This is an open source
project, so you are most welcome to contribute. Note the code quality
and design are quite important and many people will see and review
your contributions.
These changes can be done by the addition of new terms into the
equations for the relaxation rates. The description of the relaxation
due to the fully anisotropic CST can be written in concordance with the
article by Goldman M, J magn. Reson 60, 437-452 (1984). The CST tensor
is splitted into two axially symmetric subtensors which both contribute
to the relaxation. Such approach has advantage that the spectral density
function do not change the form used in Relax program. The calculation
is done in the same manner using only different input orientation. The
function for chi^2 calculation and minimalization procedure is not
affected.
This sounds quite promising for easy integration into relax.
Definition additional input information:
The orientation of the CS tensor is defined in the new input file
containing for each studied nucleus three Euler angles in separate
columns defining the orientation of the CS tensor with respect to the
PDB frame (these values are then stored in the variable called CSEA).
Separate file contains three columns with the eigenvalues of the
chemical shielding tensor (these values are then stored in the variable
called CST).
For this part, we need to design a new user function. We'll see how
you have this designed, but the best would be to have the user
function called something like cs_tensor.read() to read values from a
file. As there are 3 geometric parameters (tensor eigenvalues) and 3
orientational parameters (Euler angles), I would recommend the
following. There should be 2 keyword arguments for this function -
'param' and 'value_col'. Then to read from 2 separate files, you
could have something like param=['alpha', 'beta', 'gamma'],
value_col=[2, 3, 4]. Therefore this gives one user function which
handles both files, or one file with everything, or 6 separate files.
Importantly this gives the user flexibility. An additional function
such as cs_tensor.set() would allow greater flexibility in scripting
for the user.
The choice of atom responsible for the significant dipole-dipole
interaction with the studied nuclei is done in the PDB file. Atoms,
which are to be considered in the calculations, should be marked in the
PDB file by adding the distance to the studied nuclei (in the 1e-10m
unit) after the z coordinate. Other atoms should have zero distance
instead. So far only atoms from the same residuum may be taken into the
account.
We'll again have to see how you have this one implemented. A function
could possibly be written to find all atoms within a sphere of fixed
size. The 1.3 line is far better for handling structural data now.
Suggested changes in the code:
generic_fns/nuclei.py added function for setting gyromagnetic ratios
for selected atoms Y (atoms dipole-dipole interaction should be
considered in the calculation) and ratio of gyromagnetic ratios of atoms
Y and X (nucleus which relaxation is studied)
This function no longer exists and the gyromagnetic ratios are
automatically determined in the 1.3 line.
generic_fns/pdb.py added function for calculating XY unit vector
from the structure
This is now a series of modules located in generic_fns/structure/.
This function already exists, although it may need to be modified for
your purposes.
generic_fns/runs.py added 'mf_csa' model name
The 'run' concept no longer exists, it has been replaced by the data
pipe. See the thread starting at
https://mail.gna.org/public/relax-devel/2006-10/msg00053.html
(Message-id: <1160550133.9523.54.camel@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx>).
This post lead to the creation of the 1.3 line.
prompt/interpreter and other files in prompt directory added code
for accessing Csa_data and Model_free_csa functions
The design in here has also changed. Actually, there are not many
lines which match between the 1.2 and 1.3 lines anymore.
prompt/model_free_csa.py added csa extended model_free code
(according to model_free.py)
We'll see how this develops, but I would suggest making this
independent of model-free analysis. Then it could be used in SRLS
theory, or any other analysis that can benefit from an ellipsoidal CS
tensor. It might be possible to make this automatic so the user does
nothing.
prompt/csa_data.py Class for manipulating CST and CSEA csa data.
I would suggest the name cs_tensor. relax already has the user
function classes diffusion_tensor and align_tensor. Hence the name
cs_tensor fits better into relax.
specific_fns/model_free_csa.py added csa extended model_free code
(according to model_free.py)
This could go into the model-free specific code rather than being a
separate module.
specific_fns/csa_data.py Class for manipulating CST and CSEA csa data.
This is not analysis specific, it can be used by numerous theories,
and therefore would be better placed in generic_fns/cs_tensor.py
Instead of the only jw_mf.py file in the original version of the Relax
program we added files:
jw_mf_csa1.py (calculate the spectral density function for the first
CS subtensor)
jw_mf_csa2.py (calculate the spectral density function for the
second CS subtensor)
jw_mf_csacross.py (calculate the cross correlation spectral density
function between both CS subtensors)
jw_mf_dipY.py (calculate the vector of the spectral density
functions for the dipole-dipole interactions to all nuclei Y, i.e. each
component of vector correspond to the individual dipole-dipole interaction)
direction_cosine_csa.py (calculate direction cosines of the
principal axis of the two CS pseudo tensors and first and second
derivations of the directions cosines with respect to the angles
defining the orientation of the diffusion tensor)
direction_cosine_dipY.py (calculate direction cosines of the
principal axis of the dipole-dipole interactions to atoms Y and first
and second derivations of the directions cosines with respect to the
angles defining the orientation of the diffusion tensor, data are store
as a vector, in which each component correspond to the individual
dipole-dipole interaction)
weights_csa1.py (calculate the coefficient necessary to calculate
the spectral density function for the first CS subtensor)
weights_csa2.py (calculate the coefficient necessary to calculate
the spectral density function for the second CS subtensor)
weights_csaC.py (calculate the coefficient necessary to calculate
the cross correlation spectral density function of the first and second
CS subtensor ... the form of the equation is slightly different to previous)
weights_dipY.py (calculate the coefficient necessary to calculate
the spectral density function for the first CS subtensor)
These shouldn't be a problem as they don't affect the rest of relax.
mf_csa.py (analogy of mf.py, redirect the calculation according to
the setup and initialize all necessary parameter)
Depending on how complicated this is, I would prefer the code go into
mf.py. This is still the Lipari and Szabo model-free analysis.
ri_comps_csa_dipY.py (analogy to ri_comps.py, prepare the linear
combination of the spectral density functions and the constants
corresponding to the each type of the relaxation mechanism
i.e. instead of only
data.dip_jw_comps_func[i] ("i" goes over residues)
data.csa_jw_comps_func[i] ("i" goes over residues)
is necessary to introduce:
data.dip_jw_comps_func[i] ("i" goes over residues)
data.dipY_jw_comps_func[j][i] ("i" goes over residues, "j" over
atoms Y interacting with atom X)
data.csa1_jw_comps_func[i] ("i" goes over residues)
data.csa2_jw_comps_func[i] ("i" goes over residues)
data.csaC_jw_comps_func[i] ("i" goes over residues)
and similarly for constants:
data.dip_const_func by function comp_dip_const_func
data.dipY_const_func[i] by function comp_dipY_const_func ("j" over
atoms Y interacting with atom X)
data.csa1_const_func[i] by function comp_csa1_const_func ("i" goes
over spectrometer frequencies)
data.csa2_const_func[i] by function comp_csa2_const_func ("i" goes
over spectrometer frequencies)
data.csaC_const_func[i] by function comp_csaC_const_func ("i" goes
over spectrometer frequencies)
similarly for gradients and Hessian. So far the fitting the distance to
the selected neighbouring nuclei and the fitting of parameters of CS
tensor is not included.
This can all go into mf.py with careful design.
ri_prime_csa_dipY.py (analogy of ri_prime.py, but relaxation rates,
gradients and Hessians comprises all terms calculated by mf_csa.py )
ri_csa_dipY.py (analogy to ri.py, but again the number of variables
is enlarged by those introduced previously)
All comments or suggestions are welcomed.
Pavel Kaderavek, Petr Novak
I hope we can get your code into relax without too much pain and then
make it extremely easy for the end user to use the ellipsoidal CS
tensor and neighbouring dipoles for studying the dynamics of RNA and
other systems. This will take work to migrate to the 1.3 line though.
Cheers,
Edward
Re: relax_csa