Wednesday, 11 April 2012

unstable refinement error in SHELXL

From: Lu Yu
Date: 19 March 2012 16:00


Hi all,

I was using SHELXL for the refinement of a small peptide molecule (6-7 residues), and it was working for the first round. But then it gave me an error message. I don't know what's going on and have you had the same problems? Can you give me some suggestions?

For more information:
I was using coot to read in .fcf and .res file, and after model building, coot can generate an .ins file. I was using this .ins file and the original .hkl for the next round of SHELXL, except I added one line "ANIS" in the .ins file:

"DEFS 0.02 0.1 0.01 0.04
CGLS 10 -1
SHEL 10 0.1
FMAP 2
PLAN 200 2.3
LIST 6
WPDB 2
ANIS"

I checked the working .ins and not-working (generated from coot) .ins files,
1) the working .ins (generated from .res file at the very beginning) has:

"WGHT    0.100000
SWAT    1.352762    2.1931
FVAR   2.6206  0.5  0.5  0.5  0.5"

2)  the not-workind(generated from coot) .ins has:

"WGHT  0.1
FVAR  1.00000"
 
The refinement is shown as follows:
 "Read instructions and data
 ** Warning: unusual EXTI or SWAT parameter **
 ** Warning:    8 bad CHIV instructions ignored **
 Data:    6342 unique,      0 suppressed   R(int) = 0.0000   R(sigma) = 0.0615
 Systematic absence violations:    0    Bad equivalents:    0
 wR2 =  0.4370 before cycle   1 for   6058 data and  1690 /  1690 parameters
 GooF = S =     4.279;     Restrained GooF =      5.862  for   2243 restraints
 Max. shift = 0.259 A for O_1131b    Max. dU =-0.409 for O_1131b   at 06:08:24
 wR2 =  0.7365 before cycle   2 for   6058 data and  1690 /  1690 parameters
 GooF = S =    12.229;     Restrained GooF =     12.221  for   2243 restraints
 Max. shift = 0.175 A for O_2131a    Max. dU =-0.157 for O_5017    at 06:08:25

 ** REFINEMENT UNSTABLE **"



The other peptide dataset also has similar problem:
I was using coot-SHELXL, model building - refinement cycle successfully for the first 3 rounds, but then, it gave me an error message:

 "Read instructions and data
 ** Warning: unusual EXTI or SWAT parameter **
 ** Warning: no match for    1 atoms in CONN **

 ** CANNOT RESOLVE ISOR .. O    > LAST **"

I checked the working .ins and not-working  .ins files (both generated from coot this case),
1) the working .ins:
"WGHT    0.100000
SWAT    1.288932    3.0398
FVAR   2.6472  0.5  0.5"

2) the not-working .ins:
"WGHT    0.100000
SWAT    1.344708    3.0452
FVAR   2.731  0.54231  0.5409"

I was really confused, since I was using coot for model building for other datasets, and the .ins file generated from coot is good for SHELXL, but it didn't work all the time, eg. it work for the first few rounds, then there is a problem.

Can you give me some suggestions about what  I should do to get the SHELXL running again?

Thanks,
Lu


----------
From: Tim Gruene

Dear Lu Yu,

your wR2 in the first log-extract seems very high (43%) - it might
simply be that you model is still not good enough to refine the data
anisotropically.

Does it work if you refine the model isotropically? If so, improve the
model as much as possible before going anisotropically.

You might also want to include more reflections - with "SHEL 10 0.1" you
leave out all data with d>10A which might be quite a few important
reflections.

It is difficult to go into more detail without knowing the exact content
of the output (lst-files), especially in the second case where the
listing file tells you that something is wrong with the 'ISOR'-command.

Regards,
Tim



On 03/19/12 17:00, Lu Yu wrote:
> Hi all,
>
> I was using SHELXL for the refinement of a small peptide molecule (6-7
> residues), and it was working for the first round. But then it gave me an
> error message. I don't know what's going on and have you had the same
> problems? Can you give me some suggestions?
>
> *For more information*:
> I was using coot to read in .fcf and .res file, and after model building,
> coot can generate an .ins file. I was using this .ins file and the original
> .hkl for the next round of SHELXL, except I added one line "ANIS" in the
> .ins file:
>
> "DEFS 0.02 0.1 0.01 0.04
> CGLS 10 -1
> SHEL 10 0.1
> FMAP 2
> PLAN 200 2.3
> LIST 6
> WPDB 2
> *ANIS*"
>
> I checked the working .ins and not-working (generated from coot) .ins
> files,
> 1)* the working .ins* (generated from .res file at the very beginning) has:
>
> "WGHT    0.100000
> SWAT    1.352762    2.1931
> FVAR   2.6206  0.5  0.5  0.5  0.5"
>
> 2)  *the not-workind(generated from coot) .ins* has:
>
> "WGHT  0.1
> FVAR  1.00000"
>
> *The refinement is shown as follows:*
>  "Read instructions and data
>  ** Warning: unusual EXTI or SWAT parameter **
>  ** Warning:    8 bad CHIV instructions ignored **
>  Data:    6342 unique,      0 suppressed   R(int) = 0.0000   R(sigma) =
> 0.0615
>  Systematic absence violations:    0    Bad equivalents:    0
>  wR2 =  0.4370 before cycle   1 for   6058 data and  1690 /  1690 parameters
>  GooF = S =     4.279;     Restrained GooF =      5.862  for   2243
> restraints
>  Max. shift = 0.259 A for O_1131b    Max. dU =-0.409 for O_1131b   at
> 06:08:24
>  wR2 =  0.7365 before cycle   2 for   6058 data and  1690 /  1690 parameters
>  GooF = S =    12.229;     Restrained GooF =     12.221  for   2243
> restraints
>  Max. shift = 0.175 A for O_2131a    Max. dU =-0.157 for O_5017    at
> 06:08:25
>
>  ** REFINEMENT UNSTABLE **"
>
>
>
> The other peptide dataset also has similar problem:
> I was using coot-SHELXL, model building - refinement cycle *successfully
> for the first 3 rounds*, but then, it gave me an error message:
>
>  "Read instructions and data
>  ** Warning: unusual EXTI or SWAT parameter **
>  ** Warning: no match for    1 atoms in CONN **
>
>  ** CANNOT RESOLVE ISOR .. O    > LAST **"
>
> I checked the working .ins and not-working  .ins files (both generated from
> coot this case),
> 1) the *working .ins*:
> "WGHT    0.100000
> SWAT    1.288932    3.0398
> FVAR   2.6472  0.5  0.5"
>
> 2) the* not-working .ins*:
> "WGHT    0.100000
> SWAT    1.344708    3.0452
> FVAR   2.731  0.54231  0.5409"
>
> I was really confused, since I was using coot for model building for other
> datasets, and the .ins file generated from coot is good for SHELXL, but it
> didn't work all the time, eg. it work for the first few rounds, then there
> is a problem.
>
> Can you give me some suggestions about what  I should do to get the SHELXL
> running again?
>
> Thanks,
> Lu
>

- --
- --
Dr Tim Gruene


----------
From: George Sheldrick


Dear Lu Yu,

SHELXL is usually very stable so there must be an error in your .ins file, but it is difficult fo us to guess what it is without seeing the full file. A common error that can cause such instability is caused by a long-standing bug in Coot, which sets some occupancies in the .ins file to 1.0 (meaning that they can be refined freely starting at 1.0) rather than the usual 11.0 (which means that they should be fixed at 1.0; you can add 10 to a parameter to fix it). Another possibility is that Coot has not understood a 'free variable' that has been used for e.g. occupancy refinement. The small molecule people use other graphical GUIs for SHELXL (shelXle, WinGX, Olex2, System-S, XSEED, Oscail, XSHELL etc.) that make far fewer mistakes. The .ins files written by Coot should always be checked carefully and if necessary edited before running SHELXL.

Best wishes, George
- 


No comments:

Post a Comment