DIALS meeting 2020-11-25
Adding GEMMI as dependency of DIALS
The GEMMI (https://gemmi.readthedocs.io/en/latest/) conda-forge package is now fixed, so we can start using this.
There are a few known shortcomings with PHIL.
At present, there seems to be no way to generate a PHIL
scope object from a
scope_extract object. The
scope_extract object is the thing that has parameters to which one can assign values but it does not retain type definitions. The
scope object has type definitions but we cannot assign values to it. It’s not clear to me why we have both objects. It would be nice to be able to do the equivalent of
- Parse a PHIL string → PHIL object; and
- assign values to parameters of the PHIL object, subject to them matching the parameter type definitions; and
- add and remove parameters of the PHIL object, modify type definitions, etc.; and
- create a PHIL string representation of a PHIL object for saving to a
Naturally, if we can do all those things, we would probably also like to be able to create a PHIL object instance from scratch and add a tree of parameter definitions programatically, rather than having to work from a PHIL string as a starting point.
- possibly want a flag type:
"x *y *z"+
"*x *y *z"and not
"*x y z"
- lack of typing
- lack of useful
- can’t be used inside DUI
- Open questions about licensing raised
- JBE: has been investigating I23 anomalous datasets which don’t solve with dials compared to xds. Various attempts at reprocessing unsuccessful, even when cc_anom, dI/s(I) improved. One issue identified (likely more of an issue for I23 datasets than anomalous in general) is that you get a small percentage (<1%) of groups where outlier rejection giving wrong results - excluding too many. Cause; high intensities, high I/sigma, but high variation within group, so z-score based outlier approach not really suitable for these groups. GW: What is physical underlying cause? Not integrating some reflections well, missing something in modelling?
- JBE: Proposed that bad groups issue could be dealt with in outlier rejection. Identify bad groups, then either exclude the whole group, or keep all and use ‘internal variance’ to give these more realistic sigmas. JBE reckons could be big problem for phasing if we’re giving bad estimates of high intensity reflections with spurious accuracy. JBE will explore to see if this can help with successfully phasing I23 datasets.
- JBE: For general anomalous signal issues, still need to do work to verify https://github.com/xia2/xia2/pull/539 before ready to go, but would be good to get this in soon.
- Outcomes from last collaboration call with RJR etc.