Map Comparisons -- Am I doing something wrong?

[dbuckley-git]

Hello,

I have successfully run Misalignment Correction and subsequently Anisotropy Correction + Relion post-processing on my dataset (same dataset from my previous thread; ~1.5 mil particles, ~2.1 A reconstruction from CryoSPARC).

I've attached a PNG comparing three maps:

• my best CryoSPARC map
• the final output from Relion 3D autorefine with Misalignment Correction (run_class001.mrc)
• and the result of submitting the final Relion half-maps to another round of Anisotropy Correction and subsequent Relion post-processing

This was my command for Misalignment Correction (I had to run a 'continue' job as this dataset is quite large. Many of the flags were marked as invalid Relion options, but I assume they were retained, right?):

orterun -np 5 which relion_refine_mpi --continue Refine3D/job005/run_it012_optimiser.star --o Refine3D/job005/run --auto_refine --split_random_halves --i J490_particles_exported_processed_mrcs.star --ref cryosparc_P8_J490_008_volume_map_sharp.mrc --firstiter_cc --ini_high 30 --dont_combine_weights_via_disc --scratch_dir /local/scratch/$USER/$SLURM_JOB_ID --pool 30 --pad 2 --ctf --particle_diameter 175 --flatten_solvent --zero_mask --solvent_mask 1-1_binarized_cryosparc_P8_J498_mask.mrc --solvent_correct_fsc --oversampling 1 --healpix_order 2 --auto_local_healpix_order 5 --offset_range 5 --offset_step 2 --sym C1 --low_resol_join_halves 40 --norm --scale --j 12 --gpu "" --external_reconstruct --keep_lowres --pipeline_control Refine3D/job005/

and these were the commands for Anisotropy Correction + Relion post-processing:

spisonet.py fsc3d run_half1_class001_unfil.mrc run_half2_class001_unfil.mrc 1-1_binarized_cryosparc_P8_J498_mask.mrc --ncpus 16 --limit_res 3.5

spisonet.py reconstruct run_half1_class001_unfil.mrc run_half2_class001_unfil.mrc --aniso_file FSC3D.mrc --mask 1-1_binarized_cryosparc_P8_J498_mask.mrc --limit_res 3.5 --epochs 50 --alpha 2 --beta 0.5 --output_dir isonet_maps_2 --gpuID 0,1,2,3 --acc_batches 2

relion_postprocess --i ../isonet_maps/corrected_run_half1_class001_unfil.mrc --i2 ../isonet_maps/corrected_run_half2_class001_unfil.mrc --mask ../1-1_binarized_cryosparc_P8_J498_mask.mrc --auto_bfac

I've colored an alpha-helix that had some slight streaking in my original map yellow, as well as a nearby tryptophan yellow. For the helix, it appears Misalignment Correction alone had the best result, but there is unwanted streakiness around that nearby tryptophan that was not present in my original CryoSPARC map (and the tryptophan density itself is much worse).

Additionally, while I obtained ~2.1 A from CryoSPARC, the resolution of the final map from Relion 3D autorefine with Misalignment Correction is ~2.4 A. Is that expected?

I am really excited to use spIsoNet, I am hoping this is just some error on my end with specifying certain parameters or something minor like that!

Thank you,
David