spisonet

spIsoNet Misalignment Correction Example

This example for Misalignment is largely drawn from the original spIsoNet tutorial, using data described in the manuscript hosted at Zenodo.

Downloading tutorial data

This tutorial data is ~20G compressed, and will need ~30-35G space available to run.

#!/usr/bin/env bash

wget "https://zenodo.org/records/12640059/files/Misalignment_Correction_tutorial.tar.gz"
if [ ! -d tutorial/ ] ; then
        tar zxvf Misalignment_Correction_tutorial.tar.gz
else
        echo "tutorial directory already present, not unpacking"
fi

Running the misalignment correction

This will default to using all available GPUs, but the number of MPI processes may need to be adjusted appropriately. It should be run within the tutorial directory unpacked above.

#!/usr/bin/env bash
source /programs/sbgrid.shrc

export RELION_X=4.0.2_cu12.2
export RELION_EXTERNAL_RECONSTRUCT_EXECUTABLE="relion_wrapper.py"
export CONDA_ENV="spisonet"
JOBDIR=job001

if [ ! -d Refine3D/${JOBDIR} ] ; then
        mkdir -p Refine3D/${JOBDIR}
else
        echo "job directory already present"
        echo "please update the example, remove the existing directory, or run in a different location"                                                                                 
        echo "bailing out now"
        exit 1
fi

mpirun.relion -np 3 `which relion_refine_mpi` --o Refine3D/${JOBDIR}/run --auto_refine --split_random_halves --i job025_tutorial.star --ref HA_reference.mrc --firstiter_cc --ini_high 10 --dont_combine_weights_via_disc --pool 30 --pad 2  --ctf --particle_diameter 170 --flatten_solvent --zero_mask --solvent_mask mask.mrc --solvent_correct_fsc  --oversampling 1 --healpix_order 2 --auto_local_healpix_order 5 --offset_range 5 --offset_step 2 --sym C3 --low_resol_join_halves 40 --norm --scale  --j 4 --gpu "" --external_reconstruct --keep_lowres --pipeline_control Refine3D/${JOBDIR}/