░█████╗░██╗██╗░░░██╗░██████╗███████╗███████╗██╗░░██╗███████╗██████╗░
██╔══██╗██║██║░░░██║██╔════╝██╔════╝██╔════╝██║░██╔╝██╔════╝██╔══██╗
███████║██║╚██╗░██╔╝╚█████╗░█████╗░░█████╗░░█████═╝░█████╗░░██████╔╝
██╔══██║██║░╚████╔╝░░╚═══██╗██╔══╝░░██╔══╝░░██╔═██╗░██╔══╝░░██╔══██╗
██║░░██║██║░░╚██╔╝░░██████╔╝███████╗███████╗██║░╚██╗███████╗██║░░██║
╚═╝░░╚═╝╚═╝░░░╚═╝░░░╚═════╝░╚══════╝╚══════╝╚═╝░░╚═╝╚══════╝╚═╝░░╚═╝ 

The application of next generation sequencing (NGS) in infectious disease surveillance and outbreak investigations has become a promising area. Environmental sampling provides a method to collect and identify potentially dangerous pathogens that are circulating in an animal population, however detection of a low-abundance pathogen target in a large metagenomic background is still a challenge. AIV_seeker pipeline that is optimized for detecting and identifying low-abundance avian influenza virus (AIV) from metagenomic NGS data.

The workfow was originally built in Perl, but now we decided to switch to Nextflow-DSL2 as workflow engine starting from version 0.3. Nextflow makes the pipeline more scalable and reproducible. It's easy to run tasks across multiple compute infrastructures, and also it can support conda/Docker/Singularity containers making installation trivial and results highly reproducible.

A detailed structure and each module of the workflow is presented below in the dataflow diagram.

The pipeline currently takes reads from metagenomics sequencing as the starting point, then does a QC check. It performs quality trims the reads and adapters with fastp, and performs basic QC with FastQC and summerize the QC QC results with MultiQC. You can also get a quick taxonomic report with Krakne2 or Centrifugre by setting skip_kraken2=false or --skip_centrifuge=false.

The pipeline then preprocesses the sequences by merging pair-end reads with Vsearch: mergepairs, performing quality filtering with Vsearch: fasta_filter and dereplication with Vsearch: derep_fulllength.

After preprocessing, it performs two rounds of alignments to balance running time and accuracy. The first round is done with Diamond, then the second round is done with BLAST.

Then the alignment results are parsed. Some custom scripts are used to caculate BSR score and assign subtypes based on a custom database and metadata, which was built based on the data download from IRD.

An optional step can be done before subtyping. It uses Vsearch to group reads and then uses custom scripts to generate OTU tables and detect potential reads resulted from bleeding, which is a known issue from Illumina platform.

Furthermore, the pipeline creates reports in the results directory specified, summarizing some of the subtyping results and sequences.

1. Install nextflow (>=20.04.0)

2. Install Docker, Singularity or Conda

3. Download the pipeline and test it on a demo dataset with a single command:

   nextflow run cidgoh/AIV_seeker -profile singualrity --input demo_data/samplesheet.csv

4. Start running your own analysis!

5. Once your run has completed successfully, clean up the intermediate files.

    nextflow clean -f -k

For further information or help, don't hesitate to get in touch at [email protected] or [email protected]

An extensive list of references for the tools used by the workflow can be found in the CITATIONS.md file.