Seqtk

# 2) Subsample reads reproducibly
seqtk sample -s100 reads.fq.gz 100000 > reads.subset.fq

# 3) Extract named regions or sequences
seqtk subseq genome.fa regions.bed > subset.fa

# 4) Trim low-quality FASTQ ends
seqtk trimfq reads.fq.gz > reads.trimmed.fq

Recommended Workflow

1. Pick the exact subcommand first: seq for format conversion, sample for downsampling, subseq for extraction, trimfq for simple trimming, or fqchk for quick QC.
2. Treat seqtk as a stream-oriented filter and redirect output explicitly.
3. Use a fixed sampling seed when you need reproducible subsets.
4. Validate output record counts or names before feeding the result into heavier downstream tools.

Guardrails

• seqtk does not use global --help or --version flags; run it without arguments or use the specific subcommand help patterns instead.
• Most subcommands write to stdout by default, so forgetting redirection can make pipelines look like they succeeded without leaving files behind.
• subseq is convenient, but if you only need a few genomic intervals from a large indexed FASTA, samtools faidx may be a better fit.
• For paired-end subsampling, keep mate handling and random seed strategy consistent across both files.