scylla

This pipeline performs classification and taxonomic read binning for metagenomic reads

For each sample provided, the following analyses are performed:

• Input reads are QC filtered with fastp by length, quality, ambiguous bases and adaptors and polyX runs are trimmed (default settings)
• Read length and quality stats are collected and single read files are checked for duplicate read names
• Reads are classified using Kraken2 and the PlusPF database to get per-read taxon classifications
• For species level taxa meeting the threshold percentage and counts (higher threshold for bacteria, lower for viruses etc), a file of binned taxon reads is extracted
• The human read count in classifications is checked to make sure it is less than the threshold (1000). It is not possible to extract human-classified reads in any output.
• Reads are screened for high consequence infectious diseases (HCID) using both classified counts and read mapping to a reference genome panel (including close confounders) and checking the proportion of the genome covered.
• Reads are screened for the specified spike ins and a summary collected.
• Separate files of the unclassified read fraction, the viral+unclassified read and all non-human reads are extracted.
• A single sample report is generated

This pipeline can optionally perform a reclassification of the viral+unclassified read fraction with a custom viral database.

There is the option to perform de novo viral assembly and classification using virbot or genomad.

This pipeline can be run on a single fastq/pair of fastq files, or a directory of fastq files (which will be concatenated), or a demultiplexed directory of barcode subdirectories.

On personal computers we recommend running with the --local flag to ensure more reasonable resource requirements.

Example command

These examples use the --local flag to set resource requirements for a laptop. This will use a smaller kraken database, and therefore when working on a cluster we recommend you remove this flag.

1. Run in mSCAPE ingest mode (assumes input is a single sample, allows single/paired fastq file input).

nextflow run main.nf --fastq test/test_data/barcode01/barcode01.fq.gz -profile docker --local

or

nextflow run main.nf --fastq1 test/test_data/illumina/barcode02_R1.fq.gz --fastq2 test/test_data/illumina/barcode02_R2.fq.gz --paired -profile docker --local```

2. Run on a demultiplexed run directory (providing a run_dir). This must either contain a subdirectory per barcode, or pairs of files with names in the format *_R{1,2}*.f*q*.

nextflow run main.nf --run_dir test/test_data [--paired] -profile docker --local

3. Run a module independently of the main workflow. Supported modules are [preprocess, qc_checks, centrifuge_classification, kraken_classification, sourmash_classification, check_hcid_status, check_spike_status, extract_all, classify_novel_viruses].

nextflow run main.nf --module preprocess --fastq test/test_data/barcode01/barcode01.fq.gz -profile docker --local