Workflow description

This workflow was designed to perform analyzes of Bulk RNA-sequencing data, from FASTQ files to identification of differentially expressed genes.

It is deliberately not automated, and requires launching the scripts manually one after the other, keeping full user control and allowing custom options, whilst retaining some standardization and repeatability. Therefore, it is well suited for users who want to perform analyzes step by step by taking the time to understand the results of one step before launching the next one, and possibly change the options accordingly.

Two quantification methods are available in this workflow :

• STAR Raw Counts : This method generates a count table with integer values corresponding to exact aligned read numbers for each gene. Reads are first mapped using STAR, then quantified using featureCount.
• RSEM Estimation : This method generates a table with decimal numbers corresponding to transcript expression estimation, obtained by normalizing the mapped reads according to transcript sizes. Expression estimation is then carried out by RSEM.

Common steps

0. Preparation of references : To perform mapping to reference genome/transcriptome, it must be indexed first. To do so, it requires reference genome (FASTA file) and genome annotation (GTF file) available for download in Ensembl.org gateway.
   Note : For a quantification with featureCount genome indexing must be preformed with STAR, whereas for transcript expression estimation it must be performed with RSEM.

1. Quality Check : Quality of each FASTQ file is assessed using FastQC. A quality control report per file is then obtained, providing information on the quality of the bases, the length of the reads, the presence of adapters, etc. To make the visualization of the results easier, all reports are then pooled and analyzed simultaneously using MultiQC.

2. Trimming : According to the conclusions drawn from the quality control of the reads, a trimming step is often necessary. This step makes it possible to clean the reads, for example by eliminating sequences enriched in adapters, or by trimming poor quality bases at the ends of the reads. For this, Trimmomatic needs to be provided with the adapter sequences used for sequencing if an enrichment has been detected.
   A quality control is carried out on the FASTQ files resulting from trimming to ensure that the quality obtained is satisfactory.

STAR Raw Counts Workflow

3. Alignment to the genome : This step consists of aligning the FASTQ files to the previously indexed reference genome in order to identify the regions from which the reads come. STAR thus generates BAM files containing the reads aligned to the genome.

4. Alignment Quality Check : In order to analyze the proportion of correctly aligned reads, MultiQC can be directly used to pool the quality control of the BAM files resulting from the alignment.

5. Quantification : This step uses featureCounts to convert the BAM files containing the aligned reads into a count table usable for further analyzes in R or Python.

RSEM Estimation Workflow

3. Transcripts estimation : This step consists of aligning the FASTQ files using STAR and RSEM to make an estimate of the abundance of each transcript. Resulting .genes.results and .isoforms.results files contain respectively the results of the estimation of expression by genes or by transcripts which will be used for further analyzes in R or Python.

4. Alignment Quality Check : In order to analyze the proportion of correctly aligned reads, MultiQC can be directly used to pool the quality control of the BAM files resulting from the alignment.

Post-Processing Data Analysis

Following data analyzes are performed locally using R or Python. A complete script for basic DESeq2 analysis while performing STAR Raw Counts Workflow is provided in R script folder.

Initialization and recommendations

Scripts

All required scripts are available in the script folder in this directory.
To get more information about using these scripts, enter the command sh <script_name.sh> help.

Environments

The workflow is encoded in Shell language and is supposed to be launched under a Linux environment.
Moreover, it was written to be used on a computing cluster with tools already pre-installed in the form of modules. Modules are so loaded using module load <tool_name> command. If you use manually installed environments, simply replace module loading in script section by the environment activation command.
All script files launch tasks as qsub task submission. To successfully complete the workflow, wait for all the jobs in a step to be completed before launching the next one.

Requirments

Name                        Version
fastqc                      0.11.9
multiqc                     1.13
trimmomatic                 0.39
rsem                        1.3.2
star                        2.7.10b
subread                     2.0.1

Project diretcory

To start the workflow, create a new directory for the project and put previously downloaded scripts inside. Use it as working directory for the following steps (except for reference indexing). Create a 'Raw' subdirectory and put all the raw FASTQ files inside.
Raw FASTQ files must be compressed in '.fq.gz' or '.fastq.gz' format. If it is not the case, you need to compress them using gzip Raw/*.fastq.

For the following example, this type of folder tree is used :

Workflow Step by Step

Common Steps

0. Preparation of references

This step only needs to be carried out during the first alignment. The genome or transcriptome once indexed can be reused as a reference for subsequent alignments.
First, you need to download reference genome FASTA file and annotaion GTF file in the Genome folder.

# Example with mouse genome from Ensembl.org
wget https://ftp.ensembl.org/pub/release-108/fasta/mus_musculus/dna/Mus_musculus.GRCm39.dna_sm.primary_assembly.fa.gz
wget https://ftp.ensembl.org/pub/release-108/gtf/mus_musculus/Mus_musculus.GRCm39.108.gtf.gz

Then, create a directory for the reference and use provided scripts in refindex folder of this repository according to the workflow you aim to perform.

STAR indexing

Syntax : sh STAR_refindex.sh <FASTA> <GTF>

sh STAR_refindex.sh ../Genome/Mus_musculus.GRCm39.dna_sm.primary_assembly.fa.gz ../Genome/Mus_musculus.GRCm39.108.gtf.gz

RSEM indexing

Syntax : sh RSEM_refindex.sh <FASTA> <GTF> <build_name>

sh RSEM_refindex.sh ../Genome/Mus_musculus.GRCm39.dna_sm.primary_assembly.fa.gz ../Genome/Mus_musculus.GRCm39.109.gtf.gz mm39.108

Once indexing is done, every following steps are performed directly in the project directory.

1. Quality Check

Syntax : sh QC.sh <input_dir>

sh QC.sh Raw

Pooled results are available in ./QC/MultiQC/QC_Raw_MultiQC.html file.

2. Trimming

If low quality bases or adapter enrichment is detected, you will need to perform trimming step.
Provided trimming script allow several options :

• -S (Slingdingwindow) : Perform a sliding window trimming, cutting once the average quality within the window falls below a threshold.
• -L (Leading) : Remove low quality bases from the beginning.
• -T (Trailing) : Remove low quality bases from the end.
• -M (Minlen) : This module removes reads that fall below the specified minimal length.
• -I (Illuminaclip) : Cuts adapters and other Illumina-specific sequences present in the reads.

For more details, please read Trimmomatic Manual.

Syntax : sh Trim.sh [options] <SE|PE> <input_dir>

sh Trim.sh -S 4:15 -L 3 -T 3 -M 36 -I ../Ref/Trimmomatic/NexteraPE-PE.fa:2:30:10 PE Raw

Perform a quality check after trimming to ensure all adapters and low quality bases have been removed correctly.

sh QC.sh Trimmed/Trimmomatic/Paired

STAR Raw Counts

3. Alignment to genome

Syntax : sh STAR.sh <SE|PE> <input_dir> <refindex>

sh STAR.sh PE Trimmed/Trimmomatic/Paired ../Ref/refdata-STAR-mm39.108/GenomeDir

4. Quality Check

Syntax : sh QC.sh <input_dir>

sh QC.sh STAR

Pooled results are available in ./QC/MultiQC/STAR_MultiQC.html file.

5. Quantification

Syntax : sh Count.sh <SE|PE> <input_dir> <GTF>

sh Count.sh PE STAR ../Ref/Genome/Mus_musculus.GRCm39.108.gtf

RSEM Estimation

3. Transcripts Estimation

Because RSEM will generate specific files after the estimation of the expression, it is possible to ignore output BAM files generation using -B false option. This could be usefull to avoid saturation of disk storage space.

Syntax : sh RSEM.sh [options] <SE|PE> <input_dir> <refindex>

sh RSEM.sh -B false PE Trimmed/Trimmomatic/Paired ../Ref/refdata-RSEM-mm39.108/mm39.108

4. Quality Check

Syntax : sh QC.sh <input_dir>

sh QC.sh RSEM

Pooled results are available in ./QC/MultiQC/RSEM_MultiQC.html file.