Skip to content

CLI Reference

Complete documentation for all DeepChopper command-line interface commands.

Reference guide covering all available commands, options, and usage examples.

Overview

DeepChopper offers several commands for processing Nanopore direct-RNA sequencing data:

  • deepchopper predict - Detect adapter sequences in FASTQ files
  • deepchopper chop - Remove detected adapter sequences from reads
  • deepchopper web - Launch interactive web interface
  • deepchopper --version - Display version information

Global Options

These options are available for all commands:

deepchopper [OPTIONS] COMMAND
Option Description
--help, -h Show help message and exit
--version Show version and exit

Commands

predict

Detect adapter sequences and chimeric artifacts in sequencing data.

deepchopper predict [OPTIONS] DATA_PATH

Arguments

Argument Type Required Description
DATA_PATH Path Yes Path to FASTQ or Parquet file

Options

Option Short Type Default Description
--gpus -g Integer 0 Number of GPUs to use (0 for CPU)
--output -o Path None Output directory for predictions
--batch-size -b Integer 12 Batch size for processing
--workers -w Integer 0 Number of data loader workers
--model -m String rna002 Model to use (rna002 or rna004)
--limit-batches Integer None Limit number of batches to process
--max-sample Integer None Maximum number of samples to process
--verbose -v Flag False Enable verbose output

Examples

Basic prediction (CPU):

deepchopper predict raw_reads.fastq

Using GPU acceleration:

deepchopper predict raw_reads.fastq --gpus 1

Using RNA004 model:

deepchopper predict raw_reads.fastq --model rna004

Custom output location:

deepchopper predict raw_reads.fastq --output ./predictions

Process a subset for testing:

deepchopper predict raw_reads.fastq --max-sample 1000

Adjust batch size for memory:

deepchopper predict raw_reads.fastq --batch-size 32 --gpus 1

Output

Creates a directory containing:

  • predictions/ - Directory containing predicted adapter positions for each read

chop

Remove adapter sequences from reads based on predictions.

deepchopper chop [OPTIONS] PREDICTIONS_DIR FASTQ_FILE

Arguments

Argument Type Required Description
PREDICTIONS_DIR Path Yes Directory containing predictions from predict command
FASTQ_FILE Path Yes Original FASTQ file to chop

Options

Option Short Type Default Description
--output -o Path Auto Output FASTQ file path
--chunk-size Integer 10000 Number of reads to process per chunk
--min-read-length Integer 20 Minimum length of chopped reads to keep
--smooth-window Integer 21 Sliding window size for smoothing
--min-interval-size Integer 13 Minimum adapter region size
--max-process-intervals Integer 4 Maximum adapter regions per read
--verbose -v Flag False Enable verbose output

Examples

Basic chopping:

deepchopper chop predictions/ raw_reads.fastq

Custom output file:

deepchopper chop predictions/ raw_reads.fastq --output chopped.fastq

Memory-efficient processing:

deepchopper chop predictions/ raw_reads.fastq --chunk-size 1000

High-performance processing:

deepchopper chop predictions/ raw_reads.fastq --chunk-size 50000

Adjust sensitivity:

deepchopper chop predictions/ raw_reads.fastq --smooth-window 31 --min-interval-size 15

Keep only longer reads:

deepchopper chop predictions/ raw_reads.fastq --min-read-length 50

Output

Creates:

  • Chopped FASTQ file with adapter sequences removed
  • Reads split at adapter positions
  • Statistics about processing

Memory Usage Guidelines

Chunk Size Memory Usage Speed Use Case
1,000 ~1-2 GB Slower Memory-constrained systems
10,000 (default) ~5-10 GB Balanced General use
50,000 ~20-50 GB Fastest High-memory systems

web

Launch interactive web interface for DeepChopper.

deepchopper web [OPTIONS]

Options

Option Type Default Description
--port Integer 7860 Port to run the web server on
--share Flag False Create public Gradio link

Examples

Start web interface:

deepchopper web

Custom port:

deepchopper web --port 8080

Create shareable link:

deepchopper web --share

Note

The web interface is limited to processing one FASTQ record at a time. For batch processing, use the predict and chop commands.

Complete Workflow Examples

Basic Workflow

# 1. Predict adapters
deepchopper predict raw_reads.fastq --output predictions

# 2. Chop reads
deepchopper chop predictions/ raw_reads.fastq --output chopped.fastq

GPU-Accelerated Workflow

# Use GPU for faster prediction
deepchopper predict raw_reads.fastq --gpus 1 --batch-size 32 --output predictions

# Chop with default settings
deepchopper chop predictions/ raw_reads.fastq --output chopped.fastq

Memory-Efficient Workflow

# Process in smaller batches
deepchopper predict raw_reads.fastq --batch-size 8 --output predictions

# Chop with small chunks
deepchopper chop predictions/ raw_reads.fastq --chunk-size 1000 --output chopped.fastq

RNA004 Workflow

# Use RNA004 model
deepchopper predict raw_reads.fastq --model rna004 --output predictions

# Adjust parameters for RNA004
deepchopper chop predictions/ raw_reads.fastq \
  --smooth-window 31 \
  --min-interval-size 15 \
  --output chopped.fastq

Testing Workflow

# Test on subset of data
deepchopper predict raw_reads.fastq --max-sample 100 --output test_predictions

# Check results
deepchopper chop test_predictions/ raw_reads.fastq --output test_chopped.fastq

Tips and Best Practices

Performance Optimization

  1. Use GPU when available: Adds --gpus 1 for 10-50x speedup
  2. Adjust batch size: Larger batches are faster but use more memory
  3. Use appropriate chunk size: Balance between memory and speed for chopping
  4. Process in parallel: Run multiple instances on different files

Quality Control

  1. Inspect predictions: Check predictions/ directory for adapter positions
  2. Monitor statistics: Review processing logs for quality metrics
  3. Validate results: Compare input/output read counts
  4. Adjust parameters: Fine-tune based on your data characteristics

Troubleshooting

  • Out of memory: Reduce --batch-size or --chunk-size
  • Slow processing: Enable GPU with --gpus 1
  • Too many fragments: Increase --smooth-window or reduce --max-process-intervals
  • Missed adapters: Decrease --min-interval-size or --smooth-window

See Also