Skip to content

METHODS

Canonical Parameter Reference

Bioinformatics tool parameters, scientific rationale, and citations are documented in:

  • docs/nextflow/tool-parameters.md

Use that file as the single source of truth for analysis thresholds and tool options.

Resource Limit Policy

Resource governance has two layers:

  1. Nextflow requests in nextflow.config:
  2. cpus, memory, time
  3. Tool thread usage in process scripts:
  4. task.cpus

Rules:

  • Keep finite time for all heavy processes to prevent indefinite hangs.
  • Cap process CPUs via Math.min(params.max_cpu as int, <cap>).
  • Size memory/time by process class (light, mapping, sorting, SV calling).

Small Machine Debugging (resource request too high)

If the host has less RAM than requested (for example 32 GB host vs 64/96 GB request), tasks usually fail to start or are OOM-killed (exit 137), and pipeline execution stops because errorStrategy = 'terminate'.

Quick checks:

  1. data/outputs/logs/trace.tsv:
  2. look at status, exit, memory, peak_rss, realtime
  3. failing task logs:
  4. .command.err, .command.log in that task work/ directory

Typical error signatures:

  • Killed
  • Out of memory
  • Cannot allocate memory
  • exit code 137

Recovery:

  1. Run with lower global CPU pressure: --max_cpu <small_value>
  2. Lower per-process memory/time requests in nextflow.config for the target machine
  3. Resume: nextflow run main.nf ... -resume

Recommended operational setup:

  • maintain separate profiles for low-resource and high-resource environments (for example small, prokaryote, eukaryote).