Inferring and evaluating network medicine-based disease modules with nextflow.

Area of Science:

• Network Medicine
• Computational Biology
• Bioinformatics

Background:

• Complex diseases arise from intricate gene and protein interactions.
• Network-based methods identify disease mechanisms by expanding seed genes into disease modules.
• Divergent algorithmic strategies in module detection tools complicate the selection of biologically plausible and useful modules.

Purpose of the Study:

• To develop and present an integrated pipeline for the systematic evaluation of disease module detection tools.
• To provide guidance on selecting appropriate algorithms and networks for disease module discovery.
• To promote reproducible research in network medicine.

Main Methods:

• Implementation of an all-in-one Nextflow pipeline for tool installation, data preparation, execution, and evaluation.
• Systematic assessment of six module detection algorithms across 50 disease-network combinations.
• Evaluation criteria included module topology, functional coherence, robustness, and seed recovery.

Main Results:

• Substantial variability observed in disease modules derived from different networks and algorithms.
• Module detection methods demonstrate robustness to minor perturbations but struggle with recovering omitted seeds.
• No single method consistently outperformed others, emphasizing the importance of context-specific tool selection.

Conclusions:

• The developed pipeline facilitates systematic comparison of disease module discovery approaches.
• Findings underscore the need for careful consideration of network and algorithm choices in network medicine.
• The pipeline is integrated into nf-core, serving as an extendable resource for the research community.