4.5.7. RNA-seq

current version:0.19.0

https://badge.fury.io/py/sequana-rnaseq.svg JOSS (journal of open source software) DOI https://github.com/sequana/rnaseq/actions/workflows/main.yml/badge.svg

This is is the rnaseq pipeline from the Sequana project


RNASeq analysis from raw data to feature counts


A set of Fastq Files and genome reference and annotation.


MultiQC reports and feature Counts




Cokelaer et al, (2017), ‘Sequana’: a Set of Snakemake NGS pipelines, Journal of Open Source Software, 2(16), 352, JOSS DOI doi:10.21105/joss.00352



You must install Sequana first:

pip install sequana

Then, just install this package:

pip install sequana_rnaseq

For all dependencies (see hereafter), you can use conda. Another experimental solution is to use damona:

pip install damona
damona install sequana_tools

This will install all required dependencies.


sequana_pipelines_rnaseq --help
sequana_pipelines_rnaseq --input-directory DATAPATH --genome-directory genome --aligner star

This creates a directory with the pipeline and configuration file. You will then need to execute the pipeline:

cd rnaseq
sh rnaseq.sh  # for a local run

This launch a snakemake pipeline. If you are familiar with snakemake, you can retrieve the pipeline itself and its configuration files and then execute the pipeline yourself with specific parameters:

snakemake -s rnaseq.rules -c config.yaml --cores 4 --stats stats.txt

Or use sequanix interface.


This pipelines requires lots of third-party executable(s). Here is a list that may change. A Message will inform you would you be missing an executable:

  • bowtie
  • bowtie2
  • STAR
  • featureCounts (subread package)
  • picard
  • multiqc

You can install most of the tools using damona:

damona create --name sequana_tools
damona activate sequana_tools
damona install sequana_tools

Or use the conda.yaml file available in this repository. If you start a new environment from scratch, those commands will create the environment and install all dependencies for you:

conda create --name sequana_env python 3.7.3
conda activate sequana_env
conda install -c anaconda qt pyqt>5
pip install sequana
pip install sequana_rnaseq
conda install --file https://raw.githubusercontent.com/sequana/rnaseq/master/conda.yaml


This pipeline runs a RNA-seq analysis of sequencing data. It runs in parallel on a set of input FastQ files (paired or not). A brief HTML report is produced together with a MultiQC report.

This pipeline is complex and requires some expertise for the interpretation. Many online-resources are available and should help you deciphering the output.

Yet, it should be quite straigtforward to execute it as shown above. The pipeline uses bowtie1 to look for ribosomal contamination (rRNA). Then, it cleans the data with cutapdat if you say so (your data may already be pre-processed). If no adapters are provided (default), reads are trimmed for low quality bases only. Then, mapping is performed with standard mappers such as star or bowtie2 (--aligner option). Finally, feature counts are extracted from the previously generated BAM files. We guess the strand and save the feature counts into the directoy ./rnadiff/feature_counts.

The pipelines stops there. However, RNA-seq analysis are followed by a different analysis (DGE hereafter). Although the DGE is not part of the pipeline, you can performed it with standard tools using the data in ./rnadiff directory. One such tool is provided within our framework (based on the well known DEseq2 software).

Using our framework:

cd rnadiff
sequana rnadiff --design design.csv --features all_features.out --annotation ANNOT \
       --feature-name FEAT --attribute-name ATTR

where ANNOT is the annotation file of your analysis, FEAT and ATTR the attribute and feature used in your analysis (coming from the annotation file).

This produces a HTML repot summarizing you differential analysis.

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.


the RNAseQC rule is switch off and is not currently functional in version 0.9.X


Version Description
  • Fix bowtie2 rule to use new wrappers. Use wrappers in add_read_group and mark_duplicates
  • Adapt to new bowtie2 align wrapper
  • fix feature counts plots missing in multiqc results
  • fix regression bug introduced in snakemake 6.9.0
  • Allow the aligners to have dedicated index for each version in the same genome directory.
  • Ribosomal is now estimated on the first 100,000 reads to speed up analysis
  • --indexing and --force-indexing options not required anymore. Indexing will be done automatically and not redone if present.
  • Use of the new sequana-wrappers repository
  • Major update to use the new sequana version and the RNADiff tools.
  • remove fastq_screen. One can use sequana_multitax for taxonomic content and contamination.
  • cutadapt is now replaced by fastp, although it can still be used.
  • full integration of salmon for prokaryotes and eukaryotes
  • user interface has now a --skip-gff-check option. Better handling of input gff with more meaningful messages
  • integration of rseqc tool
  • indexing was always set to True in the config after 0.9.16 update.
  • BUG fix: Switch mark_duplicates correctly beore feature counts
  • rnadiff one factor is simplified
  • When initiating the pipeline, provide information about the GFF
  • mark duplicates off by default
  • feature_counts has more options in the help. split options into feature/attribute/extra_attributes.
  • HTML reports better strand picture and information about rRNA
  • refactorising the main standalone and config file to split feature counts optiions into feature and attribute. Sanoty checks are ow provided (--feature-counts-attribute, --feature-counts-feature-type)
  • can provide a custom GFF not in the genome directory
  • can provide several feature from the GFF. Then, a custom GFF is created and used
  • fix the --do-igvtools and --do-bam-coverage with better doc
  • 9/12/2020
  • Fixed bug in sequana/star_indexing for small genomes (v0.9.7). Changed the rnaseq requirements to benefit from this bug-fix that could lead to seg fault with star aligner for small genomes.
  • Report improved with strand guess and plot
  • 7/12/2020
  • BUG in sequana/star rules v0.9.6. Fixed in this release.
  • In config file, bowtie section 'do' option is removed. This is now set automatically if rRNA_feature or rRNA_file is provided. This allows us to skip the rRNA mapping entirely if needed.
  • fastq_screen should be functional. Default behaviour is off. If set only phiX174 will be search for. Users should build their own configuration file.
  • star/bowtie1/bowtie2 have now their own sub-directories in the genome directory.
  • added --run option to start pipeline automatically (if you know what you are doing)
  • rnadiff option has now a default value (one_factor)
  • add strandness plot in the HTML summary page
  • Remove the try/except around tolerance (guess of strandness) to make sure this is provided by the user. Final onsuccess benefits from faster GFF function (sequana 0.9.4)
  • Fix typo (regression bug) + add tolerance in schema + generic title in multiqc_config. (oct 2020)
  • add the tolerance parameter in the feature_counts rule as a user parameter (config and pipeline).
  • Best feature_counts is now saved into rnadiff/feature_counts directory and rnadiff scripts have been updated accordingly
  • the most probable feature count option is now computed more effectivily and incorporated inside the Snakemake pipeline (not in the onsuccess) so that multiqc picks the best one (not the 3 results)
  • the target.txt file can be generated inside the pipeline if user fill the rnadiff/conditions section in the config file
  • indexing options are filled automatically when calling sequana_rnaseq based on the presence/absence of the index of the aligner being used.
  • salmon now integrated and feature counts created (still WIP in sequana)
  • FastQC on raw data skipped by default (FastQC for processed data is still available)
  • Added paired options (-p) for featureCounts
  • Switch back markduplicates to False for now.
  • Use only R1 with bowtie1
  • set the memory requirements for mark_duplicates in cluster_config file
  • Set temporary directory for mark_duplicates to be local ./tmp
  • set mark_duplicate to true by default
  • use new sequana pipeline manager
  • export all features counts in a single file
  • custom HTML report
  • faster --help calls
  • --from-project option added
  • include salmon tool as an alternative to star/bowtie2
  • include rnadiff directory with required input for Differential analysis
  • Automatic guessing of the strandness of the experiment
  • Fix multiqc for RNAseQC rule
  • Fix RNAseQC rule, which is now available.
  • Fix ability to use existing rRNA file as input
  • Fix indexing for bowtie1 to not be done if aligner is different
  • add new options: --feature-counts-options and --do-rnaseq-qc, --rRNA-feature
  • Based on the input GFF, we now check the validity of the rRNA feature and feature counts options to check whether the feature exists in the GFF
  • schema is now used to check the config file values
  • add a data test for testing and documentation
  • fix typo found in version 0.9.6
  • Fixed empty read tag in the configuration file
  • Possiblity to switch off cutadapt section
  • Fixing bowtie2 rule in sequana and update the pipeline accordingly
  • Include a schema file
  • output-directory parameter renamed into output_directory (multiqc section)
  • handle stdout correctly in fastqc, bowtie1, bowtie2 rules
0.9.3 if a fastq_screen.conf is provided, we switch the fastqc_screen section ON automatically

Major refactorisation.

  • remove sartools, kraken rules.
  • Indexing is now optional and can be set in the configuration.
  • Configuration file is simplified with a general section to enter the genome location and aligner.
  • Fixed rules in sequana (0.8.0) that were not up-to-date with several executables used in the pipeline including picard, fastq_screen, etc. See Sequana Changelog for details with respect to rules changes.
  • Copying the feature counts in main directory ready to use for a differential analysis.