Hi Elaine, Oliver, Thanks for passing this on. The difference is indeed as Tristan pointed out that the sampling of points around an atom is a bit different in Chimera and ChimeraX. Tristan picked a faster method to do this in ChimeraX, whereas the original implementation uses a slower search that aims to make the points more evenly spread around each atom, while a) keeping the number of points the same at each radial distance and b) making sure the points are closer to the atom in question than to any other atom in the model. In my tests, the difference is not huge between Chimera and ChimeraX Q-scores, about 1% on average, but there was one case reported where it was ~5%, this should be more rare. Also please make sure to use sigma of 0.4 in the ChimeraX tool, if you are comparing to EMDB values. Initially we were using 0.6 which is the default there, but had to decrease it to 0.4 after the highest resolutions reached ~1Å. The value of 0.6 was based on the highest resolution being ~1.5Å. MapQ was updated the default to 0.4, which is what EMDB is also using. This will make a much larger difference in the Q-scores, ~10-15%. I haven’t looked in detail at the jamaliki code, but again differences could be to similar issue. Paul Emsley implemented a version in Coot, if that is of use, which actually seemed a lot more consistent with the MapQ calculations. I am planning to work on the ChimeraX Q-score plugin, but have been slow to get to it unfortunately. I am also planning to do a stand-alone python library, which will be easier for EMDB to keep in their pipeline. Stay tuned for these, esp. the latter, in the next month or so (if all goes well). Greg P.
On Jan 26, 2026, at 11:33 AM, Elaine Meng via ChimeraX-users <chimerax-users@cgl.ucsf.edu> wrote:
Forwarding in case someone can help with the MapQ-speciic parts of this question. Thanks Elaine
Begin forwarded message:
Subject: [Chimera] Chimera qScore Date: January 25, 2026 at 10:23:00 PM PST Cc: oliver.schanz@boehringer-ingelheim.com
Dear Chimera team, We are currently aligning our in‑house cryo‑EM structure validation with community standards and encountered an issue with reproducing Q‑scores: The wwPDB references the original Chimera/MapQ implementation, but as Chimera is being phased out in favor of ChimeraX, we can no longer reproduce the original Q‑scores in ChimeraX. We have already evaluated available alternatives: ChimeraX Q‑score plugin: We observe systematic differences between Q‑scores from ChimeraX and values in wwPDB/EMDB reports. The plugin author (Tristan Croll) indicated that differences likely stem from sampling not matching the original implementation and possibly slightly different averaging. The tool has also not been actively maintained for some time. Python implementations (e.g., jamaliki/qscore and an internal prototype): These likewise yield different Q‑scores on the same model–map pairs, likely for similar reasons. Given this, we would appreciate your guidance on two points: Are there plans to migrate the original Chimera/MapQ Q‑score implementation into ChimeraX? If not (or not soon), would it be possible to obtain free or reduced‑cost access to Chimera solely for the purpose of Q‑score calculations? We are otherwise fully using ChimeraX in our workflow. Thank you very much for your feedback. Best regards, Oliver Schanz
Oliver Schanz IT Research, Development & Medicine
Boehringer Ingelheim GmbH Birkendorfer Str. 65 | 88397 Biberach
T +49 (7351) 54-7299 <tel:+49 (7351) 54-7299> M +49 (151) 15026220 <tel:+49 (151) 15026220> E oliver.schanz@boehringer-ingelheim.com <mailto:oliver.schanz@boehringer-ingelheim.com>
_______________________________________________ ChimeraX-users mailing list -- chimerax-users@cgl.ucsf.edu To unsubscribe send an email to chimerax-users-leave@cgl.ucsf.edu Archives: https://mail.cgl.ucsf.edu/mailman/archives/list/chimerax-users@cgl.ucsf.edu/