Short Reads: Subsampling Mode

Short Reads: Subsampling Mode#

The subsampling mode allows you to downsample sequencing reads before assembly, reducing computational requirements and enabling quick pipeline validation on large datasets.

Overview#

Subsampling mode is useful when:

  • You have very large datasets and want to reduce computational time and memory usage

  • You want to quickly validate the pipeline workflow before running on full datasets

  • You need to test different parameters on a subset of data

  • Computational resources are limited

When subsampling is enabled, the pipeline uses seqtk sample to randomly downsample reads from each input file before assembly. This happens per-sample, so each sample is independently subsampled.

Activation#

Note

Before running the pipeline, make sure you have cloned the repository. See Nextflow Pipeline: Usage for installation instructions.

To enable subsampling mode, use the --subsample flag along with --type shortreads (assuming you are in the dbcan-nf directory):

nextflow run main.nf \
  --input samplesheet.csv \
  --outdir results \
  --type shortreads \
  --subsample \
  -profile docker
  --skip_kraken_extraction # based on the database size of kraken2, you can skip this step if the database is too large.

Parameters#

Subsampling Parameters#

Parameter

Type

Default

Description

--subsample

boolean

false

Enable subsampling mode. Must be used with --type shortreads.

--subsample_size

integer

20000000

Number of reads per file to retain. Applied to each FASTQ file independently.

Usage Examples#

Basic Subsampling#

Subsample to 20 million reads per file (default):

nextflow run main.nf \
  --input samplesheet.csv \
  --outdir results \
  --type shortreads \
  --subsample \
  -profile docker \
  --skip_kraken_extraction # based on the database size of kraken2, you can skip this step if the database is too large.

Custom Subsampling Size#

Subsample to 5 million reads per file:

nextflow run main.nf \
  --input samplesheet.csv \
  --outdir results \
  --type shortreads \
  --subsample \
  --subsample_size 5000000 \
  -profile docker

Subsampling with Other Options#

Combine subsampling with other parameters:

nextflow run main.nf \
  --input samplesheet.csv \
  --outdir results \
  --type shortreads \
  --subsample \
  --subsample_size 10000000 \
  --skip_kraken_extraction \
  -profile docker

Behavior#

How Subsampling Works#

  1. Per-File Subsampling: Each FASTQ file (R1 and R2 for paired-end) is independently subsampled

  2. Random Sampling: Uses seqtk sample with random seed for reproducible results

  3. Preserves Pairing: For paired-end data, both R1 and R2 files are subsampled to maintain pairing

  4. Sample Naming: Subsampled samples are renamed with _subsample suffix (e.g., sample1_dna_subsample)

Limitations#

  • Mutually Exclusive: Cannot be used together with --coassembly

  • No RNA-seq: RNA-seq processing is automatically disabled when subsampling is enabled

  • Reduced Coverage: Subsampling reduces sequencing depth, which may affect assembly quality and gene detection

Output Files#

Output files follow the same structure as standard short reads mode, but with modified sample names:

  • Sample IDs are appended with _subsample suffix

  • All downstream files (assembly, annotation, abundance) use the subsampled sample names

  • Output directory structure remains the same as Short Reads Analysis Mode

Example Output Structure#

results/
├── megahit/
│   └── sample1_dna_subsample_contigs.fa.gz
├── pyrodigal/
│   ├── sample1_dna_subsample.faa.gz
│   └── sample1_dna_subsample.gff.gz
├── rundbcan/
│   └── sample1_dna_subsample_dbcan/
└── ...

Best Practices#

  1. Start Small: Begin with a small subsample size (e.g., 5-10 million reads) to validate the pipeline

  2. Scale Up: Gradually increase subsample size to find the optimal balance between quality and resources

  3. Compare Results: Compare subsampled results with full dataset to assess impact on downstream analysis

  4. Resource Planning: Use subsampling to estimate resource requirements for full dataset analysis

When to Use Subsampling#

Recommended: - Initial pipeline validation - Parameter optimization - Resource-limited environments - Very large datasets (>100M reads per sample)

Not Recommended: - Final production analysis (use full dataset) - Low-coverage samples (subsampling may further reduce coverage) - When maximum sensitivity is required

Example Results#

For example visualizations from subsampling mode, see the subsampling results section.

See Also#

Contents