Custom steps¶

Project-specific downstream analyses are declared in the config and run by the pipeline as first-class rules, with their inputs wired to the pipeline's outputs and their resources routed through the cluster profile.

custom:
  - name: fork_probs                       # rule name and output directory
    scope: sample                          # sample (default) | project
    requires: [dnascent_bam]               # final_bam is always there; add modkit_pileup, read_summary, dnascent_bam
    command: >
      python scripts/custom/fork_probs.py --bam {dnascent_bam} --reference {reference}
        --out {outdir}/{sample}.fork_probs.tsv.gz --threads {threads}
    outputs: ["{sample}.fork_probs.tsv.gz"]  # checked after the command; missing = failure
    threads: 4
    mem_mb: 16000
    prefix: ""                             # e.g. "module load R/4.5.1 &&"

  - name: cohort_table
    scope: project                         # runs once, after every sample's final BAM
    command: "python scripts/custom/cohort.py --bams {bams} --out {outdir}/cohort.tsv"
    outputs: ["cohort.tsv"]

Placeholders:

scope	placeholders
sample	`{sample} {bam} {bai} {reference} {outdir} {threads} {pipeline_dir} {results_dir}` and, when required, `{pileup}` `{read_summary}` `{dnascent_bam}`
project	`{samples} {bams} {outdir} {threads} {pipeline_dir} {results_dir}`

Outputs land in output_directory/summary/custom/<name>/<sample>/ (or .../project/), with a .done sentinel and a log under logs/custom/. The command runs from the pipeline directory, so a script under scripts/custom/ can be named relatively. Commands run inside the pixi environment; anything else goes in prefix.

Cluster resources come from the profile's custom_sample_step / custom_project_step entries (walltime, partition) and from the step's own threads / mem_mb.

The test config exercises two trivial steps; see config/config-test.yml.