"Dataflow variables are spectacularly expressive in concurrent programming"
Henri E. Bal , Jennifer G. Steiner , Andrew S. Tanenbaum
With the arise of big data, techniques to analyse and run experiments on large datasets are increasingly necessary.
Parallelization and distributed computing are the best ways to tackle this kind of problem, but the tools commonly available to the bioinformaticians community, traditionally lack good support for these techniques, or provide a model that fits badly with the specific requirements in the bioinformatics domain and, most of the time, require the knowledge of complex tools or low-level APIs.
Nextflow framework is based on the dataflow programming model, which greatly simplifies writing parallel and distributed pipelines without adding unnecessary complexity and letting you concentrate on the flow of data, i.e. the functional logic of the application/algorithm.
It doesn't aim to be another pipeline scripting language yet, but it is built around the idea that the Linux platform is the lingua franca of data science, since it provides many simple command line and scripting tools, which by themselves are powerful, but when chained together facilitate complex data manipulations.
In practice, this means that a Nextflow script is defined by composing many different processes. Each process can be written in any scripting language that can be executed by the Linux platform (BASH, Perl, Ruby, Python, etc), to which is added the ability to coordinate and synchronize the processes execution by simply specifying their inputs and outputs.
Nextflow does not require any installation procedure, just download the distribution package by copying and pasting this command in your terminal:
curl -fsSL get.nextflow.io | bash
It creates the nextflow
executable file in the current directory. You may want to move it to a folder accessible from your $PATH
.
Create a file named hello.nf
with the following content and copy it to the path where you downloaded the Nextflow package.
process sayHello {
"""
printf 'Hello world! \n'
"""
}
Launch the above example by typing the following command on your terminal console:
./nextflow run -process.echo true hello.nf
Congratulations! You have just run your first program with Nextflow.
Let's see a more real example: execute a BLAST search, get the top 10 hits, extract the found protein sequences and align them.
Copy the following example into a file named pipeline.nf
.
params.query = "$HOME/sample.fa"
params.db = "$HOME/tools/blast-db/pdb/pdb"
process blast {
output:
file top_hits
"""
blastp -query ${params.query} -db ${params.db} -outfmt 6 \
| head -n 10 \
| cut -f 2 > top_hits
"""
}
process extract {
input:
file top_hits
output:
file sequences
"blastdbcmd -db ${params.db} -entry_batch $top_hits > sequences"
}
process align {
input:
file sequences
echo true
"t_coffee $sequences 2>&- | tee align_result"
}
The input
and output
declarations in each process, define what it is expecting to receive as input and what file(s)
are going to be produced as output.
Since the two variables query
and db
are prefixed by the params
qualifier, their values can be overridden quickly
when the script is launched, by simply adding them on the Nextflow command line and prefixing them with the --
characters.
For example:
./nextflow run pipeline.nf --db=/path/to/blast/db --query=/path/to/query.fasta
Processes in your pipeline can be written in any scripting language supported by the underlying Linux platform. To use a scripting other than Linux BASH (e.g. Perl, Python, Ruby, R, etc), simply start your process script with the corresponding shebang declaration. For example:
process perlStuff {
"""
#!/usr/bin/env perl
print 'Hi there!' . '\n';
"""
}
process pyStuff {
"""
#!/usr/bin/env python
x = 'Hello'
y = 'world!'
print "%s - %s" % (x,y)
"""
}
Nextflow provides an abstraction between the pipeline functional logic and the underlying processing system. Thus it is possible to write your pipeline once and have it running on your computer or a cluster resource manager without modifying it.
Currently the following clusters are supported:
- Open Grid Engine (SGE)
- Univa Grid Engine
- IBM Platform LSF
- Linux SLURM
- PBS/Torque
- HTCondor (experimental)
By default processes are parallelized by spanning multiple threads in the machine where the pipeline is launched.
To submit the execution to a SGE cluster create a file named nextflow.config
, in the directory
where the pipeline is going to be launched, with the following content:
process {
executor='sge'
queue='<your execution queue>'
}
In doing that, processes will be executed as SGE jobs by using the qsub
command, and so your pipeline will behave like any
other SGE job script, with the benefit that Nextflow will automatically and transparently manage the processes
synchronisation, file(s) staging/un-staging, etc.
Alternatively the same declaration can be defined in the file $HOME/.nextflow/config
, which is supposed to hold
the global Nextflow configuration.
Nextflow provides out of the box support for the Amazon AWS cloud allowing you to setup a computing cluster, deploy it and run your pipeline in the AWS infrastructure in a few commands.
The cloud configuration settings need to be specified in the nextflow.config
file as shown below:
cloud {
imageId = 'ami-43f49030'
instanceType = 't2.micro'
subnetId = 'subnet-05222a43'
sharedStorageId = 'fs-1803efd1'
spotPrice = 0.04
}
aws {
accessKey = 'xxx'
secretKey = 'yyy'
region = 'eu-west-1'
}
Replace the settings in the above example with values of your choice. The attribute sharedStorageId
is optional,
when provided the Amazon EFS file system is automatically mounted in the configured
cloud environment. The spotPrice
attribute allows you to use EC2 Spot instances
in place of regular on-request instances, bidding for the specified price.
The settings in the aws
block can be omitted, in that case Nextflow will use the AWS credentials defined in
your environment, using the standard AWS variables and configuration files.
Once defined the configuration of your cloud environment, run the following command in the folder where the file
nextflow.config
was created:
nextflow cloud create my-cluster -c <num-of-nodes>
The string my-cluster
identifies the cluster instance. Replace it with a name of your choice.
Finally replace num-of-nodes
with the actual number of instances that will made-up the cluster.
WARNING: you will be charged accordingly the type and the number of instances chosen.
Once the cluster deployment completes, SSH in the master node following the instruction that will be printed. Then you will be able to run your Nextflow pipeline as usual.
Java 7 or higher
Nextflow is written in Groovy (a scripting language for the JVM). A pre-compiled, ready-to-run, package is available at the Github releases page, thus it is not necessary to compile it in order to use it.
If you are interested in modifying the source code, or contributing to the project, it worth knowing that the build process is based on the Gradle build automation system.
You can compile Nextflow by typing the following command in the project home directory on your computer:
make compile
The very first time you run it, it will automatically download all the libraries required by the build process. It may take some minutes to complete.
When complete, execute the program by using the launch.sh
script in the project directory.
The self-contained runnable Nextflow packages can be created by using the following command:
make pack
In order to install the compiled packages use the following command:
make install
Then you will be able to run nextflow using the nextflow
launcher script in the project root folder.
Nextflow can be compiled with both JDK 7 or JDK 8. However when compiling with the later you won't be able
to run it on Java 7. The Java compiler used by the build process can be choose by setting the
JAVA_HOME
environment variable accordingly.
If the compilation stops reporting the error: java.lang.VerifyError: Bad <init> method call from inside of a branch
,
this is due to a bug affecting the following Java JDK:
- 1.7.0 update 55
- 1.7.0 update 65
- 1.7.0 update 67
- 1.7.0 update 71
- 1.8.0 update 11
- 1.8.0 update 20
Upgrade to a newer JDK to avoid to this issue. Alternatively a possible workaround is to define the following variable in your environment:
_JAVA_OPTIONS='-Xverify:none'
Read more at these links:
Nextflow development with IntelliJ IDEA requires the latest version of the IDE (14).
If you have it installed in your computer, follow the steps below in order to use it with Nextflow:
- Clone the Nextflow repository to a directory in your computer.
- Open IntelliJ IDEA and choose "Import project" in the "File" menu bar.
- Select the Nextflow project root directory in your computer and click "OK".
- Then, choose the "Gradle" item in the "external module" list and click on "Next" button.
- Confirm the default import options and click on "Finish" to finalize the project configuration.
- When the import process complete, select the "Project structure" command in the "File" menu bar.
- In the showed dialog click on the "Project" item in the list of the left, and make sure that the "Project SDK" choice on the right contains Java 7 (or higher).
Nextflow documentation is available at this link http://docs.nextflow.io
You can post questions, or report problems by using the Nextflow Google group available at this link https://groups.google.com/forum/#!forum/nextflow
The Nextflow framework is released under the GNU GPLv3 License.
If you use Nextflow for research purpose, please cite:
P. Di Tommaso, et al. Nextflow enables reproducible computational workflows. Nature Biotechnology 35, 316–319 (2017) doi:10.1038/nbt.3820
Nextflow is built on two great pieces of open source software, namely Groovy and Gpars
YourKit is kindly supporting this open source project with its full-featured Java Profiler. Read more http://www.yourkit.com