CWLab - An open-source framework for simplified deployment of the Common Workflow Language using a graphical web interface

Background and Scope:

The Common Workflow Language (CWL) allows to wrap and link up bioinformatic software in a standardized and portable way. However, setting up and operating a CWL-based workflow management system can be a labor-intensive challenge for many data-driven laboratories. To this end, we developed CWLab: a framework for simplified, graphical deployment of CWL.

CWLab allows life-science researchers with all levels of computational proficiency to create, execute and monitor jobs for CWL-wrapped tools and workflows. Input parameters for large sample batches are specified using a simple HTML form or a spreadsheet and are automatically validated. The integrated webserver allows to remotely control the execution on clusters as well as single workstations. Moreover, automatic infrastructure provisioning and scaling for OpenStack-based clouds is being implemented. CWLab can also be used as a local desktop application that supports Linux, MacOS, and Windows by leveraging Docker containerization. Our Python-based framework is easy to set up and, via a flexible API, it can be integrated with any CWL runner and adapted to custom software environments.

With CWLab, we would like to hide the complexity of workflow management so that scientific users can focus on their data analyses. This might promote the adoption of CWL in multi-professional life-science laboratories.

Installation and Quick Start:

Attention: CWLab is in alpha state currently and not all features are available yet. However, the core functionalities are working and we are happy if you test it. We are working hard to push out a stable version in the coming weeks. Please press the watch button to not miss it.

Installation can be done using pip:
python3 -m pip install cwlab

Please see the section "Configuration" for a discussion of available options.

Start the webserver with your custom configuration (or leave out the --config flag to use the default one):
cwlab up --config config.yaml

If you like to make use of containers for dependency management, you need to install Docker or a Docker-compatible containerization solution like singularity or udocker. To run on Windows or MacOs, please install the dedicated docker versions: Docker for Windows, Docker for Mac

The usage of the web interface should be self-explanatory with build-in instruction. The following section gives an overview of the basic usage scenario.

Supported Systems:

CWLab is written in platform-agnostic python and can therefore be executed on:

• Linux
• MacOs
• Windows*

Any CWL runner that has a command-line interface can be integrated into CWLab in order to execute CWL workflows or tool-wrappers, such as:

• cwltool (the reference implementation) - https://github.com/common-workflow-language/cwltool
• Toil (UCSC) - https://github.com/DataBiosphere/toil
• Cromwell (Broad Institute) - https://github.com/broadinstitute/cromwell
• Reana (CERN) - https://reana.readthedocs.io/en/latest/index.html
• CWLEXEC (IBM) - https://github.com/IBMSpectrumComputing/cwlexec (Please find a constantly updated list at: https://www.commonwl.org/#Implementations)

Therefore, CWLab can be used on any infrastructure supported by these CWL runners, including:

• single workstations
• HPC clusters (PBS, LSF, slurm, ...)
• clouds (AWS, GCP, Azure, OpenStack)

*Please Note:
Execution on Windows is only supported by cwltool which talks to docker for windows. Therefore, CWL-wrapped tools and workflows which where originally designed for Linux/MacOs can be executed on Windows with a graphical interface provided by CWLab.

Usage:

Connect to the web interface:

Open a modern browser of your choice like Chrome, Firefox, Safari, or Edge (Internet Explorer might be partially incompatible).

Type in the URL of your webserver. The URL depends on your configuration:

• If the webserver is running on the same machine and uses port 5000 is used (this is the default), type: https://localhost:5000/
• If CWLab is running on a remote machine in the same network, type in the machine's IP address and the used port. For instance, if the IP address is 172.22.0.1 and port 5000 is used:https://172.22.0.1:5000/

You should see a Welcome page like this:

Import a CWL workflow or tool:

CWLab can be used to run any workflow or tool that has been wrapped using the Common Workflow Language. Of course, you can write workflows or tool wrappers yourself (we recommend rabix-composer https://rabix.io/), however, for many especially bioinformatic tasks, existing CWL solutions are publicly available. Check the CWL website as a starting point:
https://www.commonwl.org/#Repositories_of_CWL_Tools_and_Workflows.

To import a CWL document:

• Click on the button "Import CWL Workflow/Tool" in the top bar
• Choose a CWL document (workflow or tool)*
• Press the import button

The workflow will be automatically validated:

*Please note: Currently, workflows can only be imported in the "packed" format. We will add support for the unpacked format soon. To pack a CWL workflow, use:
cwltool --pack my_workflow.cwl > my_workflow_packed.cwl

Create a new Job:

To run a workflow or tool with your data, you have to create a new job. One job may contain multiple runs (for instance multiple samples or conditions). CWLab will automatically present you a list of needed input parameters. For each parameter, you can choose whether to specify it globally (all runs of a job will get the same value) or per run.

• Click on the button "Create New Job" in the top bar and select the desired CWL document in the sidebar
• Specify a descriptive job name (the job ID will be composed of the date, time, and the name)
• If the job shall contain multiple runs toggle the "runs per job" switch, then:
  • Specify run names as a comma-separated list in the dedicated text field
  • In the parameter list, select which parameters should be run-specific
• CWLab will automatically create a parameter form for you to fill in:
  • Export/download the form in the desired format
  • Open it in a spreadsheet editor (e.g. Microsoft Excel or Open Office)
  • The file may contain the following sheets (depends on the type of input parameters and your selections for "global"/"run-specific" specification):
    • global single values: parameters that take only one value and are defined globally (one for all runs)
    • run-specific single values: parameters that take only one value but are specified per run
    • global arrays: array parameters (takes a list of values) that are defined globally
    • A separate sheet will be created for each run-specific array parameter. It will be titled with the parameters name
    • config: This sheet contains configuration options that only need adaption in advanced use cases.
  • Fill in the sheet and import/upload the edited file to CWLab *
• Your parameter settings are automatically validated. (E.g. it is checked whether the specified values match the parameter's type and whether the paths of specified files or directories exist.)
• If valid, you can press the "create job" button and head over to "Job Execution & Results" in the top bar

* Please note: For specifying file or directory parameters, there are two options:

• Either specify the absolute path
• Specify a character string that can be uniquely matched to a file/directory in the default input directory (please see the INPUT_DIR parameter in the config section).

This is an example screenshot for creating a job for an ATAC-seq workflow:

Job execution:

• Click on "Job Execution & Results" in the top bar and choose the job of interest in the sidebar
• Select the runs you want to start
• Select an execution profile (see the "Configuration" for details) and press "start"
• The execution status will be displayed in the run-list
• Pressing the "DetailsResults" button will show (not implemented yet):
  • the deployed input parameter
  • execution logs (from the CWL runner)
  • a QC report
• Once finished the output can be found in the "exec" directory (set in the configuration) along with the used parameter values, CWL document, and log files

An example screenshot of the execution interface:

Configuration:

CWLab is a highly versatile package and makes almost no assumptions on your hard- and software environment used for the execution of CWL. To adapt it to your system and use case, a set of configuration options is available:

• General configs, including:
  • webserver (hosting IP address and port, remotely or locally available, login protected or not)
  • paths of working directories
• Execution profiles:
  This flexible API allows you to adapt CWLab to your local software environment and to integrate a CWL runner of your choice (such as Cwltool, Toil, or Cromwell).

All configuration options can be specified in a single YAML file which is provided to CWLab upon start:
cwlab up --config my_config.yaml

To get an example config file, run the following command:
cwlab print_config > config.yaml (or see the example below)

General Configs:

• WEB_SERVER_HOST:
  Specify the host or IP address on which the webserver shall run. Use localhost for local usage on your machine only. Use 0.0.0.0 to allow remote accessibility by other machines in the same network.
  Default: localhost

• WEB_SERVER_PORT:
  Specify the port used by the webserver.
  Default: 5000

• TEMP_DIR:
  Directory for temporary files.
  Default: a subfolder "cwlab/temp" in the home directory

• CWL_DIR:
  Directory for saving CWL documents.
  Default: a subfolder "cwlab/temp" in the home directory

• EXEC_DIR:
  Directory for saving execution data including output files.
  Default: a subfolder "cwlab/temp" in the home directory

• INPUT_DIR:
  Directory where input files are expected by default (if the full path is not specified).
  Default: a subfolder "cwlab/temp" in the home directory

• DB_DIR:
  Directory for databases.
  Default: a subfolder "cwlab/temp" in the home directory

• DEBUG:
  If set to True, the debugging mode is turned on. Do not use on production systems.
  Default: False

Exec Profiles:

This is where you configure how to execute cwl jobs on your system. A profile consists of four steps: pre_exec, exec, eval, and post_exec (only exec required, the rest is optional). For each step, you can specify commands that are executed in bash or cmd terminal.

You can define multiple execution profile as shown in the config example below. This allows frontend users to choose between different execution options (e.g. using different CWL runners, different dependency management systems, or even choose a between multiple available batch execution infrastructures like lsf, pbs, ...). For each execution profile, following configuration parameters are available (but only shell and exec is required):

• shell:
  Specify which shell to use. For Linux or MacOS use bash. For Windows, use powershell.
  Required.

• max_retries: Specify how many times the execution (all steps) is retried before marking a run as failed.

• timeout:
  For each step in the execution profile, you can set a timeout limit.
  Default:

  pre_exec: 120
  exec: 86400
  eval: 120
  post_exec: 120

• pre_exec*:
  Shell commands that are executed before the actual CWL execution. For instance to load required python/conda environments.
  Optional.

• exec*:
  Shell commands to start the CWL execution. Usually, this is only the command line to execute the CWL runner. The stdout and stderr of the CWL runner should be redirected to the predefined log file.
  Required.

• eval*:
  The exit status at the end of the exec step is automatically checked. Here you can specify shell commands to additionally evaluate the content of the execution log to determine if the execution succeeded. To communicate failure to CWLab, set the SUCCESS variable to False.
  Optional.

• post_exec*: Shell commands that are executed after exec and eval. For instance, this can be used to clean up temporary files.

* Additional notes regarding execution profile steps:

• In each step following predefined variables are available:
  • JOB_ID
  • RUN_ID (please note: is only unique within a job)
  • CWL (the path to the used CWL document)
  • RUN_YAML (the path to the YAML file containing input parameters)
  • OUTPUT_DIR (the path of the run-specific output directory)
  • LOG_FILE (the path of the log file that should receive the stdout and stderr of CWL runner)
  • SUCCESS (if set to False the run will be marked as failed and terminated)
  • PYTHON_PATH (the path to the python interpreter used to run CWLab)
• The four steps will be executed in the same shell session and therefore can be treated as one connected script. (Between the steps, CWLab communicates the status to the database allowing the user to get status notifications via the front end).
• Thus you may define your own variables that will also be available in all downstream steps.
• At the end of each step. The exit code is checked. If it is non-zero, the run will be marked as failed. Please note, if a step consists of multiple commands and an intermediate command fails, this will not be recognized by CWLab as long as the final command of the step will succeed. To manually communicate failure to CWLab, please set the SUCCESS variable to False.
• The steps are executed using pexpect (https://pexpect.readthedocs.io/en/stable/overview.html), this allows you also connect to a remote infrastructure via ssh (recommended to use an ssh key). Please be aware that the path of files or directories specified in the input parameter YAML will not be adapted to the new host. We are working on solutions to achieve an automated path correction and/or upload functionality if the execution host is not the CWLab server host.
• On Windows, please be aware that each code block (contained in {...}) has to be in one line.

Example configuration files:

Below, you can find example configurations for local execution of CWL workflows or tools with cwltool.

Linux / MacOs:

WEB_SERVER_HOST: localhost 
WEB_SERVER_PORT: 5000

DEBUG: False  

TEMP_DIR: '/home/cwlab_user/cwlab/temp'
CWL_DIR: '/home/cwlab_user/cwlab/cwl'
EXEC_DIR: '/home/cwlab_user/cwlab/exec'
INPUT_DIR: '/home/cwlab_user/cwlab/input'
DB_DIR: '/home/cwlab_user/cwlab/db'

EXEC_PROFILES:
    cwltool_local:
        shell: bash
        max_retries: 2
        timeout:
            pre_exec: 120
            exec: 86400
            eval: 120
            post_exec: 120
        exec: |
            cwltool --outdir "${OUTPUT_DIR}" "${CWL}" "${RUN_YAML}" \
                >> "${LOG_FILE}" 2>&1
        eval: | 
            LAST_LINE=$(tail -n 1 ${LOG_FILE})
            if [[ "${LAST_LINE}" == *"Final process status is success"* ]]
            then
                SUCCESS=True
            else
                SUCCESS=False
                ERR_MESSAGE="cwltool failed - ${LAST_LINE}"
            fi

Windows:

WEB_SERVER_HOST: localhost
WEB_SERVER_PORT: 5000

DEBUG: False  

TEMP_DIR: '/home/cwlab_user/cwlab/temp'
CWL_DIR: '/home/cwlab_user/cwlab/cwl'
EXEC_DIR: '/home/cwlab_user/cwlab/exec'
INPUT_DIR: '/home/cwlab_user/cwlab/input'
DB_DIR: '/home/cwlab_user/cwlab/db'

EXEC_PROFILES:
    cwltool_windows:
        shell: powershell
        max_retries: 2
        timeout:
            pre_exec: 120
            exec: 86400
            eval: 120
            post_exec: 120
        exec: |
            . "${PYTHON_PATH}" -m cwltool --debug --default-container ubuntu:16.04 --outdir "${OUTPUT_DIR}" "${CWL}" "${RUN_YAML}" > "${LOG_FILE}" 2>&1

        eval: |
            $LAST_LINES = (Get-Content -Tail 2 "${LOG_FILE}")

            if ($LAST_LINES.Contains("Final process status is success")){$SUCCESS="True"}
            else {$SUCCESS="False"; $ERR_MESSAGE = "cwltool failed - ${LAST_LINE}"}

Documentation:

Please note: A much more detailed documentation is on the way. In the meantime, please notify us if you have any questions (see the "Contact and Contribution" section). We are happy to help.

Contact and Contribution:

If you have any question or are experiencing problems with CWLab, please contact us at [email protected] or open an issue in Github.

If you would like to contribute to the development and like to extend the functionality of CWLab to meet your requirements, you are more than welcome. We will do our best to support you and your contribution will be acknowledged.

About Us:

CWLab is developed with love at the Division of Cancer Epigenomics of the German Cancer Research Center (DKFZ) in the beautiful university city of Heidelberg. We are an interdisciplinary team with wet-lab scientists and bioinformaticians working closely together. Our DNA sequencing-driven methodologies produce challenging amounts of data. CWLab helps us by giving all members of our team the ability to perform common bioinformatic analyses autonomously without having to acquire programming skills. This allows our bioinformatic staff to focus on method development and interpretation of computationally complex data.

If you like to know more about us, please visit our website https://www.dkfz.de/en/CanEpi/contact.html.

Licence:

This package is free to use and modify under the Apache 2.0 Licence.