Nextflow pipeline:

This repository contains a small Nextflow pipeline I wrote while learning to build reproducible workflows. My goal is to move the analysis steps I ran on an HPC cluster into Nextflow ultimately, so that everything can be reproduced and shared.

What it does:

samplesheet.csv --> CUTADAPT --> FASTQC -----------------+
                       |                                 |
                       +--> SPADES --> PRODIGAL          +--> MULTIQC
                       |       |                         |
                       |       +--> BOWTIE_MAPPING ------+
                       |                  ^
                       +------------------+  (trimmed reads)
                       |
                       +--> METAPHLAN  (optional: needs --metaphlan_db)

CUTADAPT trims adapters and low-quality ends from the paired reads. FASTQC makes a quality report on the tßrimmed reads. SPADES assembles the trimmed reads into contigs (metaSPAdes, --meta). BOWTIE_MAPPING maps the trimmed reads back to those contigs with Bowtie2, then sorts and indexes the alignments with samtools and counts mapped reads. PROFIGAL predicts coding protein region from metagonome MULTIQC reds the report and log files from other tools write and puts them all into on HTML page. METAPLAN is optional. To run it, download the newest metaphlan4 database. When MetaPhlAn finishes, the per-sample profiles are merged automatically into one abundance table (merged_abundance_table.txt) with merge_metaphlan_tables.py.

1. Input: a CSV with columns sample,fastq_1,fastq_2 (paired-end reads)
2. Output (under results/): trimmed reads and cutadapt logs, FastQC reports, SPAdes contigs, and sorted BAM files with indexes, flagstat and Bowtie2 logs

Run it

1. Install Nextflow and conda
2. Download two small public test samples (from nf-core/test-datasets,taxprofiler branch)
3. nextflow run . -profile conda --input data/samplesheet.csv

Tested

• Nextflow 26.04.3 on macOS with -profile conda: cutadapt, FastQC and SPAdes, bowtie2 mapping (metaSPAdes, --meta), Prodigal, and MultiQC, completed on two small public test samples (6 tasks, about 50 min, mostly building the conda environments the first time).
• Tested optional MetaPhlAn4.0.6 and metaphlan_merged.py. The Metaphlan step was also run through Slurm on an HPC cluster and successfully passed the test.

Next