-
Notifications
You must be signed in to change notification settings - Fork 46
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
dials.refine _check_scan_range #1025
Comments
specific case: xia2/xia2#381 |
Trimming the scan to where the bright spots starts seems reasonable, but
wouldn't downstream integration be affected? IE weak spots that could have
been integrated at the beginning or end of the scan would be lost? Also,
what would be the cutoff for N strong spots? Min num spots per parameter
or zero or something in between?
…On Wed, Nov 20, 2019 at 6:06 AM Markus Gerstel ***@***.***> wrote:
specific case: xia2/xia2#381 <xia2/xia2#381>
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
<#1025?email_source=notifications&email_token=ADGY5SQPFPD4DMI3JIP4KSLQUU753A5CNFSM4JPTEYK2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEESCOUQ#issuecomment-556017490>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ADGY5SR7CFFZSZHBAHT6T5TQUU753ANCNFSM4JPTEYKQ>
.
|
It is conceivable that trimming the scan would fail to integrate some extremely weak spots that would otherwise have been integrated, but refining a scan-varying model against no data is liable to produce garbage anyway (see plot above) in which case the predictions would probably miss the spots. In this case I'm not using the minimum number of spots per parameter but just looking at the phi values for the first and last strong spot in the dataset and comparing with the scan edges. |
Following discussion with @graeme-winter about this and ideas in #1014 we are moving towards
This seems a good concept to me. So I'll work on trimming within dials.refine for scan-varying refinement. @graeme-winter suggests to trim right down to the full range of observed strong spots, considering also their shoebox |
This fixes #1025 and fixes xia2/xia2#381, but I don't like the implementation within BlockCalculator and will move it.
Before this check was introduced (#778) we would occasionally get failures in scan-varying refinement for datasets where there were no reflections at the scan edges.
Now the check is there, but beamline scientists complain that DIALS does not process data that have blank images at the beginning and/or end of the scan.
I can make the check more lenient, but at some point you just start getting a dodgy refined model, which due to the smoother also begins to affect part of the scan that really do have data. Also things like the average unit cell calculated over the scan become nonsense.
Another option would be for
dials.refine
to trim the scan to match the range of strong reflections, with a border of up to 5°, say. So, for the data set with 7.25° of images with no strong spots at the startdials.refine
could warn that scan-varying refinement would be compromised and trim the first 2.25° off.I'm not sure if
dials.refine
should be modifying the scan width like this, but in a sense the whole point of the program is to modify the experimental models, so maybe that is okay?What do people think?
The text was updated successfully, but these errors were encountered: