#kraken #bioinformatics #taxonomy #fastq #metagenomics

app kractor

Extract reads from a FASTQ file based on taxonomic classification via Kraken2

1 unstable release

0.4.0 Oct 6, 2023

#91 in Biology

22 downloads per month

MIT license

568 lines

Release GitHub release (with filter) crates.io


kraken extractor

Kractor extracts sequencing reads based on taxonomic classifications obtained via Kraken2. It consumes paired or unpaired fastq[.gz/.bz] files as input alongisde a Kraken2 standard output. It can optionally consume a Kraken2 report to extract all taxonomic parents and children of a given taxid. Fast by default, it outputs fast[q/a] files, that can optionally be compressed.

Kractor significantly enhances processing speed compared to KrakenTools for both paired and unpaired reads. Paired reads are processed approximately 21x quicker for compressed fastqs and 10x quicker for uncompressed. Unpaired reads are approximately 4x faster for both compressed and uncompressed inputs.

For additional details, refer to the benchmarks


Heavily inspired by the great KrakenTools.

At the time of writing KrakenTools operates as a single-threaded Python implementation which poses limitations in speed when processing large, paired-end fastq files. The main motivation was to enchance speed when parsing and extracting (writing) a large volume of reads - and also to learn rust!



Precompiled binaries for Linux, MacOS and Windows are attached to the latest release 0.4.0


Requires cargo

cargo install kractor

Build from source:

Install rust toolchain:

To install please refer to the rust documentation: docs

Clone the repository:

git clone https://github.com/Sam-Sims/Kractor

Build and add to path:

cd Kractor
cargo build --release
export PATH=$PATH:$(pwd)/target/release

All executables will be in the directory Kractor/target/release.


Alt text

Basic Usage:

kractor -k <kraken_output> -i <fastq_file> -t <taxonomic_id> -o <output_file> > kractor_report.json

Or, if you have paired-end illumina reads:

kractor -k <kraken_output> -i <R1_fastq_file> -i <R2_fastq_file> -t <taxonomic_id> -o <R1_output_file> -o <R2_output_file>

If you want to extract all children of a taxon:

kractor -k <kraken_output> -r <kraken_report> -i <fastq_file> -t <taxonomic_id> --children -o <output_file>




-i, --input

This option will specify the input files containing the reads you want to extract from. They can be compressed - (gz, bz2). Paired end reads can be specified by:

Using --input twice: -i <R1_fastq_file> -i <R2_fastq_file>

Using --input once but passing both files: -i <R1_fastq_file> <R2_fastq_file>

This means that bash wildcard expansion works: -i *.fastq


-o, --output

This option will specify the output files containing the extracted reads. The order of the output files is assumed to be the same as the input.

By default the compression will be inferred from the output file extension for supported file types (gz, bz). If the output type cannot be inferred, plaintext will be output.

Kraken Output

-k, --kraken

This option will specify the path to the Kraken2 output containing taxonomic classification of read IDs.


-t, --taxid

This option will specify the taxon ID for reads you want to extract.


Output type

-O, --output-type

This option will manually set the compression mode used for the output file and will override the type inferred from the output path.

Valid values are:

  • gz to output gz
  • bz2 to output bz2
  • none to not apply compresison

Compression level

-l, --level

This option will set the compression level to use if compressing the output. Should be a value between 1-9 with 1 being the fastest but largest file size and 9 is for slowest, but best file size. By default this is set at 2 as it is a good trade off for speed/filesize.

Output fasta


This option will output a fasta file, with read ids as headers.

Kraken Report

-r, --report

This option specifies the path to the report file generated by Kraken2. If you want to use --parents or --children then is argument is required.



This will extract reads classified at all taxons between the root and the specified --taxid.



This will extract all the reads classified as decendents or subtaxa of --taxid (Including the taxid).



This will output every read except those matching the taxid. Works with --parents and --children

Skip report


This will skip the json report that is output to stdout upon programme completion.

Future plans

  • Support unzipped fastq files
  • Support paired end FASTQ files
  • --include-parents and --include-children arguments
  • Supply multiple taxonomic IDs to extract
  • Exclude taxonomic IDs
  • --compression-mode
  • More verbose output
  • Benchmarks
  • Output fasta
  • Output non gz
  • Tests


  • 0.4.0



  • Json report including in stdout upon successful completion (can be disabled with --no-json)
  • Renamed


  • Support for paired-end files
  • Major under the hood changes to use Noodles for fastq parsing, and Niffler to handle compression (RIP my own code)
  • Output a fasta file with --output-fasta
  • Streamline arguments related to compression types/level
  • Improved logging


  • Code optimisations


  • Increased verbosity of outputs to user
  • --no-compress flag to output a standard, plaintext fastq file
  • --exclude to exclude specified reads. Works with --children and --parents
  • Docstrings under the hood


  • Fixes to reduce memory usage


  • Detect and handle gz files or plain files
  • --compression arg to select compression type
  • zlib-ng to speed up gzip handling
  • --children and --parents to save children and parents based on kraken report


  • First release


~179K SLoC