ESRF ID29 Bandwidth Benchmarks

Beamlines ESRF ID29
1

ID29 has two modes, 0.4% and 1.0% bandwidth. There is some uncertainty about which mode is better for data quality and, to what extent, this is due to software vs sample. Naively, the 1.0% should be better because there should be more reflections per image and more reflections should be fully recorded. We want to see if this is true in practice using both partialiator and abismal.

To this end, the ID29 team have recorded data sets of 3 samples in both modes.

Sample PDB Space Group Cell Notes Path
CuNiR 6QWG P 21 3 97.6 97.6 97.6 90.0 90.0 90.0 Anomalous, Indexing ambiguity /gpfs/ga/data/visitor/mx2731/id29/20260717/bandwidth_data/CuNiR
Lysozyme 9FUD P 43 21 2 78.7 78.7 37.86 90.0 90.0 90.0 /gpfs/ga/data/visitor/mx2731/id29/20260717/bandwidth_data/lysozyme
Myglobin 9T6V P 21 21 21 29.106 35.825 125.324 90.00 90.00 90.00 Anomalous /gpfs/ga/data/visitor/mx2731/id29/20260717/bandwidth_data/myoglobin

The plan is to merge these all in abismal and partialator and compare the results between the bandwidth settings. Two of the samples ought to have measurable anomalous signal that can be used to gauge the precision of processing.

2

Looking forward to the result!

3

Hi, just to say I am interested in this too from the DIALS side! So I will try to join in if you don’t mind? I’ll get in touch with Daniele, I did have access to ESRF at some point…

4

Happy to work with you, @JamesBE-DLS! It’s a great opportunity to benchmark dials.ssx.

5

In an ancient simulation, we got a few percent of overall Rsplit between these two bandwidth values, up to 2.1 A (2.8 A at detector side edge) based on 4/mmm symmetry, 10,000 frames and vaguely plausible distributions for scaling factors:
https://journals.iucr.org/d/issues/2013/07/00/ba5187/index.html (fig 4 and surrounding text).

At the time, we commented that “The values obtained for Rsplit in this simulation are low compared with those encountered in experiments to date”. But looking at the graph today, it’s not that far off what we’d expect for a nicely-behaved sample and experiment. So I’d be interested to see what the experiment says today!