Resolution dropped to half (3.5 to 7 .1A)in relion 4 after running auto refine

Troubleshooting
1

Hi,

I have transferred a stack of particles from the local refinement job in cryosparc to relion. But when I am running the refinement in relion the resolution is dropping to half (3.5 to 7A). I was not sure what is the problem here so I imported the motion-corrected movies from Cryosparc to relion and re-extracted the particles (Unbinned). But still, there is no improvement in the final resolution of the map.

Relion Auto_refine Parameters

data_job

_rlnJobTypeLabel relion.refine3d
_rlnJobIsContinue 0
_rlnJobIsTomo 0

version 30001

data_joboptions_values

loop_
_rlnJobOptionVariable #1
_rlnJobOptionValue #2
auto_faster No
auto_local_sampling “1.8 degrees”
ctf_intact_first_peak No
do_apply_helical_symmetry Yes
do_combine_thru_disc No
do_ctf_correction Yes
do_helix No
do_local_search_helical_symmetry No
do_pad1 No
do_parallel_discio Yes
do_preread_images No
do_queue Yes
do_solvent_fsc No
do_zero_mask Yes
fn_cont “”
fn_img Import/job001/J1013.star
fn_mask “”
fn_ref AutorefineReference.mrc
gpu_ids “”
helical_nr_asu 1
helical_range_distance -1
helical_rise_inistep 0
helical_rise_initial 0
helical_rise_max 0
helical_rise_min 0
helical_tube_inner_diameter -1
helical_tube_outer_diameter -1
helical_twist_inistep 0
helical_twist_initial 0
helical_twist_max 0
helical_twist_min 0
helical_z_percentage 30
ini_high 15
keep_tilt_prior_fixed Yes
min_dedicated 32
nr_mpi 9
nr_pool 40
nr_threads 5
offset_range 5
offset_step 1
other_args “”
particle_diameter 200
qsub qsub
qsubscript qsub_old_node-relion4.sh
queuename openmpi
range_psi 10
range_rot -1
range_tilt 15
ref_correct_greyscale No
relax_sym “”
sampling “7.5 degrees”
scratch_dir “”
skip_gridding Yes
sym_name C1
use_gpu Yes

I am not sure what I am doing wrong here. I would be really grateful if anyone can suggest possible solutions to my query.

Regards,
Shashank.

1 Like
2

Do the unmasked maps out of Relion look worse than the unmasked maps out of Cryosparc? Sometimes I’ve noticed a large difference in nominal resolution despite the maps looking the same since the 2 programs deal with masking etc. differently. It’s important to look at the unsharpened maps here since sharpening/masking the Relion map in Cryosparc to eliminate that type of difference is easy (https://guide.cryosparc.com/processing-data/all-job-types-in-cryosparc/utilities/job-sharpening-tools).

For what it’s worth, I’ve sometimes had Relion outputs look worse than Cryosparc outputs (and vice versa) and have generally assumed that it has to do with differences in how well the 2 programs deal with specific datasets (as opposed to user error).

1 Like
3

In my experience Relion can be sensitive at times to the circular mask being too big. I usually measure the volume with the scale bar in chimera and add 10Å or so. I’ve also noticed weird cases where relion can’t align particles from cryosparc initially. I’ve encountered scenarios where it seems like the z axis is rotated between the extracted particles and the input reference.

What happens when you do a command line reconstruct from relion? Something like: mpirun -np 24 relion_reconstruct_mpi --i Extract/job###/particles.star --o reconstruct.mrc --ctf --mask_diameter 200

How does the refinement look if you use map that for your input reference?

Occasionally, a single class 3D classification with a T = 2 or so, followed by a refinement also helps.