Strange results in global CTF refinement

[stefanocapaldi]

Hi.
I have a particle stack imported from relion (preprocessing: relion motion correction + CTFfind CTF estimation). When I try global CTF refinement on these particles, I get very strange results:

this is for odd terms:

And this for even terms:

the imput map is good and the estimate is between 10 and 3.4 A.
It seems to me that it does not fit anything…

When performing local CTF refinement on the same particles (with the same imput volume), it seems to work as usual with local minima close to 0.

I have done the same with other stacks from Relion in other projects and all went well.
Any suggestions?
Cheers
Strefano

[wtempel]

@stefanocapaldi Please can you post the output of the command

cryosparcm cli "get_job('P99', 'J199', 'job_type', 'version', 'status',  'params_spec')"

where you replace P99, J199 with the global CTF refinement job’s project and job ID’s, respectively
and the gold standard resolution of the currently best refined reconstruction for these particles.

[stefanocapaldi]

HI!
this is the output of the command:

{‘_id’: ‘673762d1a5553846a7fcbd9c’, ‘job_type’: ‘import_particles’, ‘params_spec’: {‘blob_exists’: {‘value’: True}, ‘ctf_exists’: {‘value’: True}, ‘enable_validation’: {‘value’: True}, ‘particle_blob_path’: {‘value’: ‘/media/stefano/data1/projects/bgal/relion/Extract/Extract_ptc_from_J016_at_360_px,_no_binning/movies’}, ‘particle_meta_path’: {‘value’: ‘/media/stefano/data1/projects/bgal/relion/Extract/Extract_ptc_from_J016_at_360_px,_no_binning/particles.star’}, ‘pick_stats_exists’: {‘value’: True}}, ‘project_uid’: ‘P14’, ‘status’: ‘completed’, ‘uid’: ‘J86’, ‘version’: ‘v4.6.0’}

the resolution of the best reconsctuction with these particles in so far is 3.48 A

[wtempel]

@stefanocapaldi

‘job_type’: ‘import_particles’

Please can you also post the output of the equivalent get_job() command for the global CTF refinement job.

[stefanocapaldi]

sorry, I digitet the wrong job number…
here is the output:
{‘_id’: ‘673b47e6a5553846a7286505’, ‘job_type’: ‘ctf_refine_global’, ‘params_spec’: {‘crg_num_iters’: {‘value’: 2}}, ‘project_uid’: ‘P14’, ‘status’: ‘completed’, ‘uid’: ‘J93’, ‘version’: ‘v4.6.0’}

[Rado]

I am experiencing a similar issue.
All processing was done in CryoSPARC 4.6.2
Running a homo refine job on a node running Ubuntu 22.04, AMD Epyc, Nvidia A6000, Nvidia driver 565.57, cuda 12.7, produces aberrant CTF refine results:

An earlier job from the pipeline using a smaller box (160 pix instead of 360 pix) run fine on this node.

Running the larger box job on a node with Rocky Linux 8.5, Intel Xeon, Nvidia RTX3090, Nvidia driver 545.23, cuda 12.3, produces correct results:

I repeated the test a couple of times with the same results.

Best wishes,

Rado

[wtempel]

@Rado Please can you post

1. for the two jobs for which you posted screenshots, the outputs of these commands

   csprojectid=P99 # replace with actual project ID
   csjobid=J199 # replace with id of a job that should be running
   cryosparcm cli "get_job('$csprojectid', '$csjobid', 'job_type', 'version', 'instance_information', 'status',  'params_spec', 'input_slot_groups')"
   cryosparcm cli "get_scheduler_targets()"
   

2. did both of those jobs run within the same project, using the same storage?
   and, if particle cashing was enabled:
3. Is the cache storage shared between the two nodes?
4. What do the results look for the 360px particle job on the ubuntu node with Cache particle images on SSD disabled?