Dear Murpholino,
I have few remarks regarding the plots in proc_stats.png
- Completeness is too low.
- Something is clearly wrong with the data processed beyond 1.3 A
resolution using xia2/dials.
- It would be useful to see statistics also for low resolution shells.
- R_meas or R_pim are much better metrics than R_merge.
You can also try to run XDS manually, for instance using xdsgui.
I wouldn't attempt to run paired refinement (and make any decisions
based on it) before these issues with data processing are addressed.
Best regards,
Martin
On 30/04/2025 18:54, Gerard Bricogne wrote:
Dear Murpholino,
Although other factors may be involved, your observations may
have something to do with the fact that you are using the 20250409
build of XDS, about which we sent the following message to all users
of autoPROC yesterday:
==
Dear all,
we've now had a chance to run more tests on the current 20250409 XDS
binary: the autoPROC website at
https://www.globalphasing.com/autoproc/wiki/index.cgi?ComparisonProcessing202504
contains the Wilson plots for the same set of PDB examples used
before. You will see the effect of the different XDS binaries and why
we currently recommend the 20230630 XDS binary (or the 20250327 one)
over others or the 20250409 one, as many of the latter show physically
implausible features in their Wilson plots. Our previous findings were
communicated to Wolfgang Kabsch and we expect an updated/fixed version
soon.
Kind regards
The autoPROC developers
==
All our observations point towards the recurrence in this binary
of a problem similar to an earlier one (encountered in previous builds
but eliminated in the 20250327 build) that led to spurious positive
biases in the integrated intensities and in the I/sigI values. Yours
seem to be concordant with ours, so you could try and see what happens
if you use 20250327.
If you are interested in the full story of our assessment of
previous XDS versions, you might want to take a look at
https://www.globalphasing.com/autoproc/wiki/index.cgi?ComparisonProcessing202503
With best wishes,
Gerard.
--
On Wed, Apr 30, 2025 at 11:19:04AM -0600, Murpholino Peligro wrote:
Dear all
While processing a dataset and comparing results from XDS and DIALS, I
encountered something strange. Please find the attached plots for reference
In summary, when using xds to process the dataset, got:
- Higher R_merge values
- Lower CC_1/2 (even got some negative values)
- Higher I/sigI
Got almost the same completeness and multiplicity (this last one not shown,
sorry).
I thought that the higher R_merge mtzs would result in noisier maps and
therefore more waters would be added by phenix.refine but... got exactly
the opposite (clearly seen in combined.png).
Running the 1.00 Å resolution MTZ files through Pairef yielded these
suggested cutoffs:
-
1.45 Å for the XIA2/XDS dataset
-
1.25 Å for the XIA2/DIALS dataset
Lowest R-free is for XIA2/XDS at 1.2 Å.
This is just one half of a dataset (45 degrees in rotation).
(If you want to play with it,
https://www.dropbox.com/scl/fi/898ey1w04qvc91d701qqm/01.tar.xz?rlkey=30887xr4dqexvympjegrpr4ie&st=xchricjg&dl=0
)
Processing was done with the CCP4i2 GUI; version 2.2.3 built on 07/Feb/25;
Linux version 9.0.008.
XDS: VERSION Jan 19, 2025 BUILT=20250409 30-Apr-2025
And the quick refinement (adding anisotropic waters only if the high
resolution cutoff was 1 or 1.1 A and setting protein atoms to be
anisotropic at all resolutions) was done with the latest stable phenix
version (1.21.2; Release tag: 5419; Platform: intel-linux-2.6-x86_64)
Is this related to
https://www.globalphasing.com/autoproc/wiki/index.cgi?ComparisonProcessing202504
?
Thanks for your time.
########################################################################
To unsubscribe from the CCP4BB list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1
This message was issued to members of www.jiscmail.ac.uk/CCP4BB, a mailing list
hosted by www.jiscmail.ac.uk, terms & conditions are available at
https://www.jiscmail.ac.uk/policyandsecurity/
########################################################################
To unsubscribe from the CCP4BB list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1
This message was issued to members of www.jiscmail.ac.uk/CCP4BB, a mailing list
hosted by www.jiscmail.ac.uk, terms & conditions are available at
https://www.jiscmail.ac.uk/policyandsecurity/
########################################################################
To unsubscribe from the CCP4BB list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1
This message was issued to members of www.jiscmail.ac.uk/CCP4BB, a mailing list
hosted by www.jiscmail.ac.uk, terms & conditions are available at
https://www.jiscmail.ac.uk/policyandsecurity/
