Skip to content

code_contribution.md

Latest commit

 

History

History
151 lines (106 loc) · 7.03 KB

code_contribution.md

File metadata and controls

151 lines (106 loc) · 7.03 KB

Code Contribution Guidelines

To make the EAGER2 code and processing logic more understandable for new contributors, and to ensure quality. We are making an attempt to somewhat-standardise the way the code is written.

If you wish to contribute a new module, please use the following coding standards.

The typical workflow for adding a new module is as follows:

  1. Define the corresponding input channel into your new process from the expected previous process channel (or re-routing block, see below).
  2. Write the process block (see below).
  3. Define the output channel if needed (see below).
  4. Add any new flags/options to nextflow.config with a default (see below).
  5. Add any new flags/options to the help message (for integer/text parameters, print to help the corresponding nextflow.config parameter).
  6. Add sanity checks for all relevant parameters.
  7. Add any new software to the scrape_software_versions.py script in bin/ and the version command to the scrape_software_versions process in main.nf.
  8. Do local tests that the new code works properly and as expected.
  9. Add a new test command in .github/workflow/ci.yaml.
  10. If applicable add a MultiQC module.
  11. Update MultiQC config assests/multiqc_config.yaml so relevant suffixes, name clean up, General Statistics Table column order, and module figures are in the right order.
  12. Add new flags/options to 'usage' documentation under docs/usage.md.
  13. Add any descriptions of MultiQC report sections and output files to docs/output.md.

Default Values

Default values should go in nextflow.config under the params scope.

Default resource processes

Defining recommended 'minimum' resource requiements (CPUs/Memory) for a process should be defined in conf/base.config. This can be utilised within the process using ${task.cpu} or ${task.memory} variables in the script: block.

Process Concept

We are providing a highly configurable pipeline, with many options to turn on and off different processes in different combinations. This can make a very complex graph structure that can cause a large amount of duplicated channels coming out of every process to account for each possible combination.

The EAGER pipeline can currently be broken down into the following 'stages', where a stage is a collection of non-terminal mututally exclusive processes, which is the output of which is used for another file reporting module (but not reporting!) .

  • Input
  • Convert BAM
  • PolyG Clipping
  • AdapterRemoval
  • Mapping (either: bwa, bwamem, or circularmapper)
  • BAM Filtering
  • Deduplication (either dedup or markduplicates
  • BAM Trimming
  • PMDtools
  • Genotyping

Every step can potentially be skipped, therefore the output of a previous stage must be able to be passed to the next stage, if the given stage is not run.

To somewhat simplify this logic, we have implemented the following structure.

The concept is as follows:

  • Every 'stage' of the pipeline (i.e. collection of mutually exclusive processes) must always have a if else statement following it.
  • This if else 'bypass' statement collects and standardises all possible input files into single channel(s) for the next stage.
  • Importantly - within the bypass statement, a channel from the previous stage's bypass mixes into these output channels. This additional channel is named ch_previousstage_for_skipcurrentstage. This contains the output from the previous stage, i.e. not the modified version from the current stage.
  • The bypass statement works as follows:
    • If the current stage is turned on: will mix the previous stage and current stage output and filter for file suffixes unique to the current stage output
    • If the current stage is turned off or skipped: will mix the previous stage and current stage output. However as there there is no files in the output channel from the current stage, no filtering is required and the files in the 'ch_XXX_for_skipXXX' stage will be used.

This ensures the same channel inputs to the next stage is 'homogenous' - i.e. all comes from the same source (the bypass statement)

An example schematic can be given as follows

 // PREVIOUS STAGE OUTPUT
if (params.run_bam_filtering) {
    ch_input_for_skipconvertbam.mix(ch_output_ch_convertbam)
        .filter{ it =~/.*converted.fq/}
        .into { ch_convertbam_for_fastp; ch_convertbam_for_skipfastp }
} else {
    ch_input_for_skipconvertbam
        .into { ch_convertbam_for_fastp; ch_convertbam_for_skipfastp }
}

// SKIPPABLE CURRENT STAGE PROCESS
process fastp {
    publishDir "${params.outdir}/fastp", mode: 'copy'

    when:
    params.run_fastp

    input:
    file fq from ch_convertbam_for_fastp

    output:
    file "*pG.fq" into ch_output_from_fastp

    script:
    """
    echo "I have been fastp'd" > ${fq}  
    mv ${fq} ${fq}.pG.fq
    """
}

// NEXT STAGE INPUT PREPARATION
if (params.run_fastp) {
    ch_convertbam_for_skipfastp.mix(ch_output_from_fastp)
        .filter { it =~/.*pG.fq/ }
        .into { ch_fastp_for_adapterremoval; ch_fastp_for_skipadapterremoval }
} else {
    ch_convertbam_for_skipfastp
        .into { ch_fastp_for_adapterremoval; ch_fastp_for_skipadapterremoval }
}

Naming Schemes

Please use the following naming schemes, to make it easy to understand what is going where.

  • process output: ch_output_from_<process>(this should always go into the bypass statement described above).
  • skipped process output: ch_<previousstage>_for_<skipprocess>(this goes out of the bypass statement described above)
  • process inputs: ch_<previousstage>_for_<process> (this goes into a process)

Nextflow Version Bumping

If you have agreement from reviewers, you may bump the 'default' minimum version of nextflow (e.g. for testing).

For this, you need to update the in the manifest{} scope of nextflow.config, and also in .travis.yml and .github/workflows/nf-core_eager.yml

Software Version Reporting

If you add a new tool to the pipeline, please ensure you add the information of the tool to the get_software_version process.

Add to the script block of the process, something like the following:

<YOUR_TOOL> --version &> v_<YOUR_TOOL>.txt 2>&1 || true

or

<YOUR_TOOL> --help | head -n 1 &> v_<YOUR_TOOL>.txt 2>&1 || true

You then need to edit the script bin/scrape_software_versions.py to

  1. add a (python) regex for your tools --version output (as in stored in the v_<YOUR_TOOL>.txt file), to ensure the version is reported as a v and the version number e.g. v2.1.1
  2. add a HTML block entry to the OrderedDict for formatting in MultiQC.

If a tool does not unfortunately offer any printing of version data, you may add this 'manually' e.g. with echo "v1.1" > v_<YOUR_TOOL>.txt

Images and Figures

For all internal nf-core/eager documentation images we are using the 'Kalam' font by the Indian Type Foundry and licensed under the Open Font License.

For the overview image we follow the nf-core style and guidelines.

These can be found for download here
⚡