QuickStart#

Easy Install#

noise2read is currently distributed on PyPI and bioconda, quick install noise2read in the build environment.

  • Install via bioconda

conda install -c bioconda noise2read

conda install py-xgboost-gpu

Note

Currently, noise2read at bioconda requires installing py-xgboost-gpu separately. I will include this dependency in later release.

  • Install via pip

pip install noise2read

and then install bioconda distributed packages of seqtk and bcool.

conda install -c bioconda seqtk bcool

Optional to install pygraphviz if you need the visualised read graph.

conda install -c conda-forge pygraphviz

Examples#

  1. A simplified version of noise2read which excludes machine learning instead uses heuristics for error correction for short reads set with default parameters.

    noise2read -m simplify_correction -i *.fa/fasta/fastq/fq -d output_directory

  2. General correction for short reads set with default parameters.

    • Training with CPU

    noise2read -m correction -i *.fa/fasta/fastq/fq -a True -d output_directory

    • Training with GPU

    noise2read -m correction -i *.fa/fasta/fastq/fq -a True -g gpu_hist -d output_directory

  3. Correcting amplicon sequencing data with default parameters

    • Training with CPU

    noise2read -m amplicon_correction -i *.fa/fasta/fastq/fq -a True -d output_directory

    • Training with GPU

    noise2read -m amplicon_correction -i *.fa/fasta/fastq/fq -a True -g gpu_hist -d output_directory

Note

We strongly recommend utilizing GPU for model training and prediction, especially for large data sets, rather than using a CPU. If a GPU resource is available; otherwise, using the simplified version of noise2read (simplify_correction) is better.