search

tristanic / isolde

Interactive molecular dynamics based model building into low-resolution crystallographic and cryo-EM maps
Other
34 stars 4 forks source link

Output of `isolde write phenixRsrInput` crashes Phenix 1.20.1 #13

Open PlethoraChutney opened 1 year ago

PlethoraChutney commented 1 year ago

Hello again! Sorry to be the bearer of bad news.

The output of isolde write phenixRsrInput crashes Phenix 1.20.1 RealSpaceRefine, during the ADP step. The issue seems to be from the generated parameter/.eff file. Using the same .cif and map, but omitting the .eff file, the RSR completes. I've tested two different map/model pairs as well. The .eff file does not crash Phenix v1.19.2

Param Value
ISOLDE Version 1.5
ChimeraX Version 1.5rc202211091945 (2022-11-09)
Phenix Versions phenix-1.19.2-4158 and phenix-1.20.1-4487
Machine (ISOLDE) M2 Macbook Air
Machines (Phenix) M2 Macbook Air (Phenix is an Intel build, so using Rosetta) and Ubuntu workstation
Map Type cryoEM
tristanic commented 1 year ago

Thanks for reporting. Could you share the text output from the Phenix crash?

PlethoraChutney commented 1 year ago

Unfortunately, it seems to crash silently. Nothing is written to the log and when running on my laptop I just got a standard "Python quit unexpectedly" error message. If you know where I might look for more logs I'd be happy to.

PlethoraChutney commented 1 year ago

The jobs never get marked as failed in Phenix's db either --- they're "running" (with no output and no PID) until manually marked as Failed

tristanic commented 1 year ago

Hmm. That sounds more like a Phenix bug to me. If I were to guess, I'd say the most likely culprit is the reference model torsion restraints (ISOLDE tells Phenix to apply them using the input model as its own reference). Your best bet would be to report this to help@phenix-online.org. They'll want a copy of the offending model, map and .eff to reproduce the problem, and will keep all your data confidential.

PlethoraChutney commented 1 year ago

OK! Thanks! Will do. Would you like me to update you on what they say?

tristanic commented 1 year ago

Absolutely - that would be great!

tristanic commented 1 year ago

I am starting to wonder if there's a hardware issue in your particular Mac. Might be worth running their diagnostic tool for peace of mind (see https://support.apple.com/en-gb/HT202731).

PlethoraChutney commented 1 year ago

Hah, that would have been funny. Just ran it and got the all-clear. I did notice something though. When opening the .cif files generated by ISOLDE, I get warnings like these:

Unknown polymer entity '1' near line 165
Unknown polymer entity '2' near line 3513
Unknown polymer entity '3' near line 7342
Unknown polymer entity '4' near line 10538
Unknown polymer entity '5' near line 11063
2 messages similar to the above omitted
Missing or incomplete entity_poly_seq table. Inferred polymer connectivity.

Always one per chain. The warnings are not present in the input .pdb models.

Could that have something to do with it? I'm noticing that Phenix's .cif files also generate a lot of warnings; is it just a more verbose format?

tristanic commented 1 year ago

Those warnings should be quite harmless. That's ChimeraX's way of saying there is no definitive sequence registered for those chains (which is the case for virtually all mmCIF files other than those downloaded directly from the PDB). The Phenix mmCIF files don't quite follow the mmCIF format specification in all details - again, mostly harmless.

Guillawme commented 9 months ago

Not sure this is helpful one year later, but for what it's worth I don't get this problem using ChimeraX 1.6.1, ISOLDE 1.6.0 and phenix 1.20.1-4487 on Rocky Linux 9 (all running on an Intel Core i7). Caveat: my sample size is only two pairs of model/map (also cryoEM maps).