nf-cosimflow

A Nextflow subworkflow to calculate cosine similarity and generate heatmaps from gene expression data.

📖 Overview

nf-cosimflow was originally designed as a standalone pipeline but has been refactored into a modular subworkflow to comply with nf-core standards.

It takes a gene expression matrix (CSV or XLSX) as input and performs the following:

1. Preprocessing: Filters genes based on mean expression (optional).
2. Calculation: Computes pairwise similarity between samples using Cosine, Pearson, or Spearman metrics.
3. Visualization: Generates a high-quality heatmap (.png) and saves the similarity matrix (.csv).

This tool is designed to be easily integrated into larger pipelines like nf-core/rnaseq or nf-core/differentialabundance for exploratory data analysis (EDA) and QC.

🚀 Features

• Modular Design: Built as a local module (compute_cosine) and subworkflow (run_cosimflow).
• Reproducible: Fully containerized environment (supports Docker and Conda).
• Flexible Inputs: Accepts standard expression matrices (genes as rows, samples as columns).
• Customizable: Choose your metric (cosine, pearson, spearman) and filtering thresholds.

🛠️ Usage

You can run the workflow directly from this repository to test it:

nextflow run main.nf \
    --input data/test_expression.csv \
    -profile docker

Parameters

Parameter	Description	Default
--input	Path to the expression matrix (CSV/XLSX).	null
--method	Similarity metric: cosine, pearson, or spearman.	cosine
--min_gene_mean	Filter out genes with mean count < value.	0.0
--outdir	Directory to save results.	results

📂 Output

All results are saved in the results/ directory:

• *_matrix.csv: The numeric similarity matrix.
• *_heatmap.png: A visual heatmap of the sample clustering.

🧩 Integration

To use this in your own Nextflow pipeline:

1. Copy modules/local/compute_cosine.nf and subworkflows/local/run_cosimflow.nf to your project.
2. Import it in your workflow:

include { RUN_COSIMFLOW } from './subworkflows/local/run_cosimflow.nf'

workflow {
    // Create input channel: [ meta, file ]
    ch_input = Channel.fromPath(params.input).map { file -> [ [id: file.baseName], file ] }
    
    RUN_COSIMFLOW(ch_input)
}

👥 Credits

Developed by Miguel Rosell (@miguelrosell). Feel free to open an issue if you find any bugs or have suggestions!