cellink.tl.external.write_slurm_array_job#

cellink.tl.external.write_slurm_array_job(adata, gwas_file, plink_bfile, prs_dir, output_dir='.', job_name='scprs_plink', ld_bfile=None, p_thresholds=None, r2_thresholds=None, clump_kb=250, maf=0.01, plink_cmd='plink', plink_threads=4, slurm_partition=None, slurm_qos=None, slurm_mem='8G', slurm_cpus=4, slurm_time='04:00:00', slurm_max_simultaneous=200, extra_sbatch_lines=None)#

Prepare per-cell GWAS files and write a SLURM array job for the PLINK C+T step.

Pre-generates per-cell filtered GWAS files in prs_dir (Python step, requires AnnData, must be run once before cluster submission).
Writes {output_dir}/{job_name}_commands.txt — one shell command per cell, where each command runs PLINK clumping + scoring for all r² thresholds for that cell.
Writes {output_dir}/{job_name}_array.sh — a SLURM array job script where task $SLURM_ARRAY_TASK_ID executes the corresponding line of the commands file.
Logs the sbatch command and the follow-up Python call to run after all jobs complete.

After all array tasks finish, load the results with:

pkl = prepare_scprs_data(
    adata=adata,
    labels=labels,
    gwas_file=gwas_file,
    prefix="my_prefix",
    prs_dir="{prs_dir}",
)

Parameters:

adata (AnnData) – Reference scATAC-seq AnnData with peak coordinates in adata.var.
gwas_file (str or Path) – Post-QC GWAS summary statistics.
plink_bfile (str or Path) – PLINK binary prefix for the target cohort (scoring).
prs_dir (str or Path) – Directory where per-cell GWAS files and PLINK score files are written. Pass this same path to prepare_scprs_data() afterwards.
output_dir (str or Path, default=".") – Where to write the commands file and SLURM script.
job_name (str, default="scprs_plink") – SLURM job name and file prefix.
ld_bfile (str or Path, optional) – PLINK prefix for LD estimation. Defaults to plink_bfile.
p_thresholds (list of float, optional) – C+T p-value thresholds.
r2_thresholds (list of float, optional) – C+T LD r² thresholds.
clump_kb (int, default=250) – Clumping window in kb.
maf (float, default=0.01) – MAF filter for PLINK scoring.
plink_cmd (str, default="plink") – PLINK executable name (plink or plink2).
plink_threads (int, default=4) – --threads passed to every PLINK call. Should equal slurm_cpus so the CPU allocation is fully utilised.
slurm_partition (str, optional) – #SBATCH --partition value (-p flag). If None, the line is omitted — useful on clusters where the default partition is fine or where -q/--qos selects the queue instead.
slurm_qos (str, optional) – #SBATCH -q value. If None, the line is omitted.
slurm_mem (str, default="8G") – Memory per task. Clumping on 1000G EUR (~8 M variants, 503 samples) needs ~4-6 GB. Increase for larger cohorts (e.g. "32G" for UKB).
slurm_cpus (int, default=4) – --cpus-per-task. Should match plink_threads.
slurm_time (str, default="04:00:00") – Wall-clock time limit per task (HH:MM:SS). Increase threads conservatively.
slurm_max_simultaneous (int, default=200) – Maximum number of concurrently running array tasks (the %N suffix on --array). Reduce to avoid overwhelming shared storage.
extra_sbatch_lines (list of str, optional) – Additional raw #SBATCH lines inserted verbatim, e.g. ["#SBATCH --nice=0", "#SBATCH --gres=gpu:0"].

Return type:

dict

Returns:

dict with keys:

commands_file: Absolute path to the commands text file.
script_file: Absolute path to the SLURM array script.
n_cells: Number of cells with at least one peak-overlapping GWAS variant (= number of array tasks).
submit_cmd: The sbatch command string to run.
next_step: Reminder string describing the Python call to make after all jobs finish.

cellink.tl.external.write_slurm_array_job

Contents

cellink.tl.external.write_slurm_array_job#