This instructional material was originally developed by Maely Gauthier in 2024 as part of the QUT eResearch infrastructure. It is free to distribute but we just require that you acknowledge eResearch for any outputs (e.g. training, presentation slides, publications) that might result from using this training material.

Some sections of this course were adapted from the Carpentry course: https://carpentries-incubator.github.io/workflows-nextflow/.

Prerequisites

You will require a basic knowledge of Linux/Unix commands to be able to participate effectively in this workshop. If you don’t, please attend the following training [Introduction to HPC].

Getting started with Nextflow

• What is a workflow and what are workflow management systems?

• Why should I use a workflow management system?

• What is Nextflow?

• What are the main features of Nextflow?

• What are the main components of a Nextflow script?

What is Nextflow?

• Nextflow is a free and open-source pipeline management software that enables scalable and reproducible scientific workflows. It allows the adaptation of pipelines written in the most common scripting languages.

• Key features of Nextflow:

  • Reproducible → version control and use of containers ensure the reproducibility of nextflow pipelines

  • Portable → compute agnostic (i.e., HPC, cloud, desktop)

  • Scalable → run from a single to thousands of samples

  • Minimal digital literacy → accessible to anyone

  • Active global community → more and more nextflow pipelines are available (i.e., https://nf-co.re/pipelines )

Nextflow is a pipeline engine that can take advantage of the batch nature of the HPC environment to efficiently and quickly run Bioinformatic workflows.

For more information about Nextflow, please visit Nextflow - A DSL for parallel and scalable computational pipelines

Installing Nextflow

Nextflow is meant to run from your home folder on a Linux machine like the HPC.

First connect to your Lyra account.

Before we start using the HPC, let’s start an interactive session:

qsub -I -S /bin/bash -l walltime=10:00:00 -l select=1:ncpus=1:mem=4gb

You should be in your home directory, if unsure you can run the following command:

cd ~

To install Nextflow, copy and paste the following block of code into your terminal (i.e., PuTTy that is already connected to the terminal) and hit 'enter':

module load java
curl -s https://get.nextflow.io | bash
mv nextflow $HOME/bin

• Line 1: The module load command is necessary to ensure java is available

• Line 2: This command downloads and assembles the parts of nextflow - this step might take some time.

• Line 3: When finished, the nextflow binary will be in the current folder so it should be moved to your “bin” folder” so it can be found later.

To verify that Nextflow is installed properly, you can run the following command:

nextflow info

We will now run locally your first Nextflow pipeline, which is called Hello:

mkdir $HOME/nftemp && cd $HOME/nftemp
nextflow run hello

• Line 1: Make a temporary folder for Nextflow to create files when it runs.

• Line 2: Verify Nextflow is working.

You should see something like this:

If you got this output, well done! You have run your first Nextflow pipeline successfully.

Now go back to your home directory and clean the test folder.

cd $HOME
rm -rf nftemp

Nextflow’s base configuration

A key Nextflow feature is the ability to decouple the workflow implementation, which describes the flow of data and operations to perform on that data, from the configuration settings required by the underlying execution platform. This enables the workflow to be portable, allowing it to run on different computational platforms such as an institutional HPC or cloud infrastructure, without needing to modify the workflow implementation.

For instance, a user can configure Nextflow so it runs the pipelines locally (i.e. on the computer where Nextflow is launched), which can be useful for developing and testing a pipeline script on your computer

\\default Nextflow settings
process {
  executor = 'local'
}

or configure Nextflow to run on a cluster such as a PBS Pro resource manager:

process {
  executor = 'pbspro'
}

The base configuration that is applied to every Nextflow workflow you run is located in $HOME/.nextflow/config.

Once you have installed Nextflow on Lyra, there are some settings that should be applied to your $HOME/.nextflow/config to take advantage of the HPC environment at QUT.

To create a suitable config file for use on the QUT HPC, copy and paste the following text into your Linux command line and hit ‘enter’. This will make the necessary changes to your local account so that Nextflow can run correctly:

[[ -d $HOME/.nextflow ]] || mkdir -p $HOME/.nextflow

cat <<EOF > $HOME/.nextflow/config
singularity {
    cacheDir = '$HOME/.nextflow/NXF_SINGULARITY_CACHEDIR'
    autoMounts = true
}
conda {
    cacheDir = '$HOME/.nextflow/NXF_CONDA_CACHEDIR'
}
process {
  executor = 'pbspro'
  scratch = false
  cleanup = false
}
includeConfig '/work/datasets/reference/nextflow/qutgenome.config'
EOF

• Line 1: Check if a .nextflow/config file already exists in your home directory. Create it if it does not exist

• Line 2-15: Using the cat command, paste text in the newly created .nextflow/config file which specifies the cache location for your singularity and conda.

• What are the parameters you are setting?

• Line 3-6 set the directory where remote Singularity images are stored and direct Nextflow to automatically mount host paths in the executed container.

• Line 7-9 set the directory where Conda environments are stored.

• Line 10-14 sets default directives for processes in your pipeline. Note that the executor is set to pbspro on line 11.

• Line 15 provides the local path to genome files required for pipelines such as nf-core/rnaseq

More in depth information on Nextflow configuration is described here: https://www.nextflow.io/docs/latest/config.html.

Nextflow pipeline repositories

nf-core

What is nf-core?

nf-core is a community-led project to develop a set of best-practice pipelines built using Nextflow workflow management system. Pipelines are governed by a set of guidelines, enforced by community code reviews and automatic code testing. The diagram below showcases the key aspects of nf-core and is divided into three sections:

• the Deploy section includes features like Stable pipelines, Centralized configs, List and update pipelines, and Download for offline us.

• the Participate section highlights Documentation, Slack workspace, Twitter updates, and Hackathons.

• the Develop section emphasizes the Starter template, Code guidelines, CI code linting and tests, and Helper tools.

What are nf-core pipelines?

nf-core pipelines are an organised collection of Nextflow scripts, other non-nextflow scripts (written in any language), configuration files, software specifications, and documentation hosted on GitHub. There is generally a single pipeline for a given data and analysis type e.g. There is a single pipeline for bulk RNA-Seq. All nf-core pipelines are open source.

Searching for available nf-core pipelines

Go to https://nf-co.re/pipelines

Narrow search by typing relevant term, for example ‘rna-seq’:

Pipelines can be sorted by Latest release, Name or Stars:

Examples of pipelines used at QUT:

• nf-core/ampliseq

• nf-core/smrnaseq

• nf-core/rnaseq

• nf-core/sarek

nf-core support

For support with Nextflow, see https://nf-co.re/join. For instance, there is a very active slack community for nf-core users.

epi2me workflows

EPI2ME Labs maintains a collection of bioinformatics workflows tailored to Oxford Nanopore Technologies long-read sequencing data. They are curated and actively maintained by experts in long-read sequence analysis.

https://eresearchqut.atlassian.net/wiki/spaces/EG/pages/edit-v2/2261090311#epi2me

Examples of pipelines used at QUT:

• wf-metagenomics

Running pipelines

Fetching pipeline code

The pull command allows you to download the latest version of a project from a GitHub repository or to update it if that repository has previously been downloaded in your home directory.

nextflow pull nf-core/<pipeline>

Please note that Nextflow would also automatically fetch the pipeline code when you run the command below for the first time:

nextflow run nf-core/<pipeline>

For reproducibility, it is good to explicitly reference the pipeline version number that you wish to use with the -revision/-r flag.

In the example below we are pulling the rnaseq pipeline version 3.12.0

nextflow pull nf-core/rnaseq -revision 3.12.0 

We can see from the output we have the latest release.

Downloaded pipeline projects are stored in the folder $HOME/.nextflow/assets.

Software requirements for pipelines

Nextflow pipeline software dependencies are specified using either Docker, Singularity or Conda. It is Nextflow that handles the downloading of containers and creation of conda environments. This is set using the -profile {docker,singularity,conda} parameter when you run Nextflow.

At QUT, we use singularity so we would specify: -profile singularity.

Install and test that the pipeline installed successfully

Pipelines generally include test code that can be run to make sure installation was successful.

From the command line

By running the Nexflow pipeline on the command line, the progress of the analysis is captured in real-time.

Run the following command from your home directory:

cd
nextflow run nf-core/smrnaseq -profile test,singularity --outdir results -r 2.1.0

This will download the smrnaseq pipeline and then run the test code. It should take ~20-30 minutes to run to completion.

It will first display the version of the pipeline which was downloaded: version 2.1.0.

It will then list all the parameters that differ from the pipeline default.

Before running a process, it will download the required singularity image.

In the screenshot below, all the jobs which will be run are listed.

We can see that 4 jobs have run to completion:

• NPUT_CHECK:SAMPLESHEET_CHECK (samplesheet.csv)

• MIRNA_QUANT:PARSE_MATURE

• MIRNA_QUANT:PARSE_HAIRPIN

• GENOME_QUANT:INDEX_GENOME (genome.fa)

One singularity image is being pulled

This is a screenshot taken half way through the analysis:

This is the output you should get when your Nextflow job has run to completion.

At the bottom, the message ‘Pipeline completed successfully’ will be printed along with the duration, the CPU hours and numbers of jobs that run to completion.

Launching Nextflow using a PBS script

Launching the Nextflow pipeline from the command line enabled us to understand what the pipeline does in real-time. But you have to make sure you keep the terminal page from which you launched the analysis opened until the analysis is done.

So now that you have learnt how to run Nextflow locally, we will use a PBS script to launch the analysis.

Move back into your home directory.

Create a test.sh script by running the following command:

cat <<EOF > $HOME/smrnaseq_test.sh
#!/bin/bash -l
#PBS -N ontvisc
#PBS -l select=1:ncpus=2:mem=8gb
#PBS -l walltime=1:00:00

cd $PBS_O_WORKDIR
module load java
NXF_OPTS='-Xms1g -Xmx4g'
nextflow run nf-core/smrnaseq -profile test,singularity --outdir results -r 2.1.0
EOF

Make the command executable and then submit your job to the PBS queue by running the following commands:

chmod +x smrnaseq_test.sh
qsub smrnaseq_test.sh

Input specifications

Samplesheet input

Nextflow pipelines generally need an input file, often referred to as a samplesheet, which contains information about the samples you would like to analyse.

The samplesheet can have as many columns as you desire, however, there is a strict requirement for the first columns to match those required by the pipeline.

The minimum information required will vary and will be specified on the usage section of the pipeline that you are interested to run.

When running Nextflow, use this parameter to specify the samplesheet location: --input '[path to samplesheet file]'

The samplesheet has to be a comma-separated file with a minimum set of columns (which will vary depending of the pipeline you are interested to run), and a header row.

Examples of samplesheets

For the nf-core/smrnaseq pipeline, the samplesheet has to be a comma-separated file with the following 2 columns.

Column names has to be specified in a header row as shown in the samplesheet example below:

sample,fastq_1
Clone1_N1,s3://ngi-igenomes/test-data/smrnaseq/C1-N1-R1_S4_L001_R1_001.fastq.gz
Clone1_N3,s3://ngi-igenomes/test-data/smrnaseq/C1-N3-R1_S6_L001_R1_001.fastq.gz
Clone9_N1,s3://ngi-igenomes/test-data/smrnaseq/C9-N1-R1_S7_L001_R1_001.fastq.gz
Clone9_N2,s3://ngi-igenomes/test-data/smrnaseq/C9-N2-R1_S8_L001_R1_001.fastq.gz
Clone9_N3,s3://ngi-igenomes/test-data/smrnaseq/C9-N3-R1_S9_L001_R1_001.fastq.gz
Control_N1,s3://ngi-igenomes/test-data/smrnaseq/Ctl-N1-R1_S1_L001_R1_001.fastq.gz
Control_N2,s3://ngi-igenomes/test-data/smrnaseq/Ctl-N2-R1_S2_L001_R1_001.fastq.gz
Control_N3,s3://ngi-igenomes/test-data/smrnaseq/Ctl-N3-R1_S3_L001_R1_001.fastq.gz

For the nf-core/rnaseq pipeline, the samplesheet has to be a comma-separated file with the following 4 columns:

Column names has to be specified in a header row as shown in the samplesheet example below:

sample,fastq_1,fastq_2,strandedness
CONTROL_REP1,AEG588A1_S1_L002_R1_001.fastq.gz,AEG588A1_S1_L002_R2_001.fastq.gz,auto
CONTROL_REP1,AEG588A1_S1_L003_R1_001.fastq.gz,AEG588A1_S1_L003_R2_001.fastq.gz,auto
CONTROL_REP1,AEG588A1_S1_L004_R1_001.fastq.gz,AEG588A1_S1_L004_R2_001.fastq.gz,auto

Please note that in this example, the same sample (CONTROL_REP1) was sequenced across 3 lanes. The nf-core/sarek pipeline will concatenate the raw reads before performing any downstream analysis.

Exercise 1

The following samplesheet file for the nf-core/rnaseq pipeline consisting of both single- and paired-end data is ready for analysis.

• How many samples does it have in total?

• How many are single-end and paired-end?

sample,fastq_1,fastq_2,strandedness
CONTROL_REP1,AEG588A1_S1_L002_R1_001.fastq.gz,AEG588A1_S1_L002_R2_001.fastq.gz,forward
CONTROL_REP2,AEG588A2_S2_L002_R1_001.fastq.gz,AEG588A2_S2_L002_R2_001.fastq.gz,forward
CONTROL_REP3,AEG588A3_S3_L002_R1_001.fastq.gz,AEG588A3_S3_L002_R2_001.fastq.gz,forward
TREATMENT_REP1,AEG588A4_S4_L003_R1_001.fastq.gz,,reverse
TREATMENT_REP2,AEG588A5_S5_L003_R1_001.fastq.gz,,reverse
TREATMENT_REP3,AEG588A6_S6_L003_R1_001.fastq.gz,,reverse
TREATMENT_REP3,AEG588A6_S6_L004_R1_001.fastq.gz,,reverse

Exercise 2

Find what are the minimal columns required in the samplesheet to run nfcore/ampliseq

Input folder

Some pipelines like nf-core/ampliseq will let you specify directly the path to the folder that contains your input FASTQ files, as an alternative to using a samplesheet.

For example:

--input_folder 'path/to/data/'

File names must follow a specific pattern, default is /*_R{1,2}_001.fastq.gz, but this can be adjusted with the --extension parameter.

For example, the following files in the folder data would be processed as sample1 and sample2:

data
    |-sample1_1_L001_R1_001.fastq.gz
    |-sample1_1_L001_R2_001.fastq.gz
    |-sample2_1_L001_R1_001.fastq.gz
    |-sample2_1_L001_R2_001.fastq.gz

All sequencing data should originate from one sequencing run, because processing relies on run-specific error models that are unreliable when data from several sequencing runs are mixed. Sequencing data originating from multiple sequencing runs requires additionally the parameter --multiple_sequencing_runs and a specific folder structure, for example:

data
    |-runA
    |   |-sample1_1_L001_R1_001.fastq.gz
    |   |-sample1_1_L001_R2_001.fastq.gz
    |   |-sample2_1_L001_R1_001.fastq.gz
    |   |-sample2_1_L001_R2_001.fastq.gz
    |
    |-runB
        |-sample3_1_L001_R1_001.fastq.gz
        |-sample3_1_L001_R2_001.fastq.gz
        |-sample4_1_L001_R1_001.fastq.gz
        |-sample4_1_L001_R2_001.fastq.gz

Where sample1 and sample2 were sequenced in one sequencing run and sample3 and sample4 in another sequencing run.

Parameters

Finding list of parameters available

For the nf-core pipelines, the tools implemented and the range of parameters available are generally described in the Usage section. Some of the parameters will be required, others optional.

Let’s have a look at the nf-core/rnaseq pipeline:

All the parameters available will also be listed under the Parameters section:

Exercise 1

Using the usage and parameters sections, search how many aligner options are available for the nf-core rnaseq pipeline version 3.14.0 .

Specifying parameters on the command line

Parameters are generally specified on the CLI (i.e. command line interface).

nextflow run nf-core/rnaseq -profile singularity -resume \
        --input samplesheet.csv \
        --outdir results \
        --genome GRCm38 \
        --aligner star_salmon \
        --extra_trimgalore_args "--quality 30 --clip_r1 10 --clip_r2 10 --three_prime_clip_r1 1 --three_prime_clip_r2 1 "

Nextflow caching

One of the core features of Nextflow is the ability to cache task executions and re-use them in subsequent runs to minimize duplicate work. Resumability is useful both for recovering from errors and for iteratively developing a pipeline

Resume option

You can enable resumability in Nextflow with the -resume flag when launching a pipeline with nextflow run.

All task executions are automatically saved to the task cache, regardless of the -resume option (so that you always have the option to resume later).

Structure of work folder

When nextflow runs, it assigns a unique ID to each task. This unique ID is used to create a separate execution directory, within the work directory, where the tasks are executed and the results stored. A task’s unique ID is generated as a 128-bit hash number.

When we resume a workflow, Nextflow uses this unique ID to check if:

1. The working directory exists

2. It contains a valid command exit status

3. It contains the expected output files.

If these conditions are satisfied, the task execution is skipped and the previously computed outputs are applied.

When a task requires recomputation, ie. the conditions above are not fulfilled, the downstream tasks are automatically invalidated.

Therefore, if you modify some parts of your script, or alter the input data using -resume, Nextflow will only execute the processes that are actually changed.

The execution of the processes that are not changed will be skipped and the cached result used instead.

This helps a lot when testing or modifying part of your pipeline without having to re-execute it from scratch.

By default the pipeline results are cached in the directory work where the pipeline is launched.

We can use the Bash tree command to list the contents of the work directory. Note: By default tree does not print hidden files (those beginning with a dot .). Use the -a to view all files.

tree -a work

Provide a relevant example from test run

Example of work directory:

work/
├── 12
│   └── 5489f3c7dbd521c0e43f43b4c1f352
│       ├── .command.begin
│       ├── .command.err
│       ├── .command.log
│       ├── .command.out
│       ├── .command.run
│       ├── .command.sh
│       ├── .exitcode
│       └── temp33_1_2.fq.gz -> /home/training/data/yeast/reads/temp33_1_2.fq.gz
├── 3b
│   └── a3fb24ad3242e4cc8e5aa0c24d174b
│       ├── .command.begin
│       ├── .command.err
│       ├── .command.log
│       ├── .command.out
│       ├── .command.run
│       ├── .command.sh
│       ├── .exitcode
│       └── temp33_2_1.fq.gz -> /home/training/data/yeast/reads/temp33_2_1.fq.gz
├── 4c
│   └── 125b5e5a5ee144fa25dd9bccd467e9
│       ├── .command.begin
│       ├── .command.err
│       ├── .command.log
│       ├── .command.out
│       ├── .command.run
│       ├── .command.sh
│       ├── .exitcode
│       └── temp33_3_1.fq.gz -> /home/training/data/yeast/reads/temp33_3_1.fq.gz
├── 54
│   └── eb9d72e9ac24af8183de569ab0b977
│       ├── .command.begin
│       ├── .command.err
│       ├── .command.log
│       ├── .command.out
│       ├── .command.run
│       ├── .command.sh
│       ├── .exitcode
│       └── temp33_2_2.fq.gz -> /home/training/data/yeast/reads/temp33_2_2.fq.gz
├── e9
│   └── 31f28c291481342cc45d4e176a200a
│       ├── .command.begin
│       ├── .command.err
│       ├── .command.log
│       ├── .command.out
│       ├── .command.run
│       ├── .command.sh
│       ├── .exitcode
│       └── temp33_1_1.fq.gz -> /home/training/data/yeast/reads/temp33_1_1.fq.gz
└── fa
    └── cd3e49b63eadd6248aa357083763c1
        ├── .command.begin
        ├── .command.err
        ├── .command.log
        ├── .command.out
        ├── .command.run
        ├── .command.sh
        ├── .exitcode
        └── temp33_3_2.fq.gz -> /home/training/data/yeast/reads/temp33_3_2.fq.gz

Task execution directory

Within the work directory there are multiple task execution directories. There is one directory for each time a process is executed. These task directories are identified by the process execution hash. For example the task directory fa/cd3e49b63eadd6248aa357083763c1 would be location for the process identified by the hash fa/cd3e49 .

The task execution directory contains:

• .command.sh: The command script.

• .command.run: The file is a bash script that Nextflow generates to execute the .command.sh script, handling the necessary environment setup and command execution details.

• .command.out: The complete job standard output.

• .command.err: The complete job standard error.

• .command.log: The wrapper execution output.

• .command.begin: A file created as soon as the job is launched.

• .exitcode: A file containing the task exit code.

• Any task input files (symlinks)

• Any task output files

Specifying another work directory

Depending on your script, this work folder can take a lot of disk space. You can specify another work directory using the command line option -w. Note Using a different work directory will mean that any jobs will need to re-run from the beginning.

Clean the work directory

If you are sure you won’t resume your pipeline execution, you can clean the work folder using the nextflow clean command. It is good practice to do so regularly.

nextflow clean [run_name|session_id] [options]

Nextflow pipeline outputs and PBS outputs

Results folder

The results are output in the folder name specified in the .nexftlow.config file under the outdir parameter. It is generally set to be results.

// nextflow.config
params {
  outdir = 'results'
}

Nextflow log, metrics and reports

By default, Nextflow will create a log file in the working directory called .nextflow.log. This file is hidden but you can see it using the command:

ls -a

If you rerun the pipeline in the same folder, the previous .nextflow.log will be renamed .nextflow.log.1 and a new .nextflow.log will be generated.

You can change the default location by specifying a different location

nextflow -log ~/code/nextflow.log run  

PBS output

Where to from now?

Nextflow offers free Fundamentals Training: https://training.nextflow.io/basic_training/

Provide links to carpentry course: https://carpentries-incubator.github.io/workflows-nextflow/instructor/01-getting-started-with-nextflow.html