Pipeline that runs bcl2fastq and creates additional plots within a Snakemake workflow

## Project description

This is is the demultiplex pipeline from the Sequana projet

Overview: Runs bcl2fastq on raw BCL data and creates plots to ease the QC validation A valid Illumina base calling directory An HTML report, a set of PNG files and the expected FastQ files production https://github.com/sequana/demultiplex/wiki This README file, the Wiki from the github repository (link above) and https://sequana.readthedocs.io Cokelaer et al, (2017), ‘Sequana’: a Set of Snakemake NGS pipelines, Journal of Open Source Software, 2(16), 352, JOSS DOI https://doi:10.21105/joss.00352

## Installation

Intall the sequana_demultiplex package as follows:

pip install sequana_demultiplex


## Usage

sequana_demultiplex --help
sequana_demultiplex --working-directory DATAPATH --bcl-directory bcldata --sample-sheet SampleSheet.csv --merging-strategy merge


The –bcl-directory option indicates where to find your raw data, the sample-sheet expects the SampleSheet to be compatible with IEM software. The –merging-strategy can be set to none or merge. The merge option merges the lanes, which is useful for e.g. NextSeq sequencers.

This creates a directory fastq. You just need to execute the pipeline:

cd demultiplex
sh demultiplex.sh  # for a local run


These commands launch a snakemake pipeline. If you are familiar with snakemake, you can retrieve the demultiplex.rules and config.yaml files and then execute the pipeline yourself with specific parameters:

snakemake -s demultiplex.rules --cores 4 --stats stats.txt \
--wrapper-prefix https://raw.githubusercontent.com/sequana/sequana-wrappers/"


You may also use sequanix for a graphical interface.

Would you need to merge the lane, please add the –merging-strategy argument followed by merge:

sequana_demultiplex --bcl-directory bcl_data --merging-strategy merge --sample-sheet SampleSheet.csv


## Requirements

This pipeline requires the following third-party tool(s):

• bcl2fastq 2.20.0

This software has an end-user license agreement (EULA). Given the EULA details of this software, it cannot be distributed by us but only by Illumina. Therefore, you should install it yourself. On cluster facility, you may ask to your system administator. For instance:

module load bcl2fastq/2.20.0


For the same reason you cannot find it on community such as bioconda or docker (aug 2020).

So, you will need to download the code yourself. The easiest is to download the RPM from Illumina and accept the agreements. Install it using the RPM if you have a debian-like system:

rpm install file.rpm


If you do not have a debian system, you can look at https://damona.readthedocs.io where we provide a singularity recipes to build an image from your own rpm. Recipes can be found here.

## Details

This pipeline runs bcl2fastq 2.20 and creates a set of diagnostics plots to help deciphering common issues such as missing index and sample sheet errors.

## Rules and configuration details

Here is the latest documented configuration file to be used with the pipeline. Each rule used in the pipeline may have a section in the configuration file.

## Changelog

Version Description
1.2.1
• Update CI action and use new sequana_pipetools v0.9.0
1.2.0
• stable release with cleanup of the setup and README
1.1.3
• add the –mars-seq option that fills the config automatically
1.1.2
• fix the none_and_force merging strategy option
1.1.1
• fix a regression bug
1.1.0
• Uses new sequana-wrappers repository
1.0.5
• Fix regression bug to cope with new snakemake API
• Compatibility with sequanix GUI
1.0.4
• Better HTML report with updated images.
• validate the SampleSheet when using sequana_demultiplex and/or the pipeline
• Add error handler from sequana_pipetools
• save all undetermined barcodes (not just first 20)
• No changes to the UI
• technically, the input_directory option is now in a section so that it can be used in Sequanix
1.0.3
• remove check_samplesheet and fix_samplesheet modules now in sequana
• check sample sheet but do not fail. Instead, informing users that there is an error and suggest to use ‘sequana samplesheet –quick-fix’
1.0.2 Use ‘sequana samplesheet –check ‘ command instead of deprecated sequana_check_sample_sheet command
1.0.1

change some default behaviour:

• write_fastq_reverse_complement is now set to False by default like bcl2fastq
• The –no-bgzf-compression option is changed into –bgzf-compression. We do not want this option by default.
• The –ignore-missing-bcls option is changed into –no-ignore-missing-bcls so as to ignore missing bcls by default keep this option as a flag and keep same behaviour
• Fix HTML syntax
1.0.0
• stable version pinned on sequana libraries
0.9.11
• fix label in plot_summary,
• add two tools one to check the samplesheet called sequana_sample_sheet and one called sequana_fix_samplesheet. The former is now inside the pipeline as well and when creating the pipeline
• set –write_reverse_complement to False by default
• remove the –ignore-missing-control which is deprecated anyway
0.9.10
• implement the new option –from-project, add missing MANIFEST
0.9.9
• simplification of the pipeline to use sequana 0.8.4 to speed up the –help calls.
• include a summary HTML report
0.9.8
• fix typos
0.9.7
• Use new release of sequana_pipetools
• set matplotlib backend to agg
• include a simple HTML report
0.9.6
• Handle different RunParameter.xml name (NextSeq vs HiSeq)
0.9.5
• Fix a regression bug due to new sequana release. We do not check the input file (fastq) since this is not a sequence analysis pipeline
• Check whether it is a NextSeq run. If so, merging-strategy must be set to ‘merge’. Can be bypassed using –force
0.9.4
• Check the presence of the bcl input directory and samplesheet.
• More help in the –help message.
• add –sample-sheet option to replace –samplesheet option
• Fix the schema file
• Check for presence of RunParameters.xml and provide information if merging-stratgy is set to None whereas it is a NextSeq run
0.9.3 Fix regression bug
0.9.2 remove warning due to relative paths.
0.9.1 Make the merging options compulsory. Users must tell whether they want to merge the lanes or not. This avoid to do the merging or not whereas the inverse was expected.
0.8.6 Uses 64G/biomics queue and 16 cores on a SLURM scheduler

## Contribute & Code of Conduct

To contribute to this project, please take a look at the Contributing Guidelines first. Please note that this project is released with a Code of Conduct. By contributing to this project, you agree to abide by its terms.

## Project details

Uploaded source