pcsf2featuremap

pcsf2featuremap provides a lightweight R interface to the core prize-collecting Steiner forest (PCSF) solver used by featuremap.

This package is designed as a minimal wrapper around the original PCSF C++ solver so that featuremap can call the core optimization routine through a small and stable interface, without depending on the full original PCSF R package and its additional visualization or enrichment-analysis dependencies.

Purpose

The original PCSF package provides a complete framework for network-based interpretation of high-throughput biological data, including PCSF optimization, randomized runs, plotting, and enrichment analysis.

For featuremap, only the core PCSF solver is required. This package exposes that solver through:

call_sr(from, to, cost, node_names, node_prizes)

featuremap builds its own metabolomics-specific workflow around this function, including:

• feature-to-metabolite network construction
• terminal prize calculation
• hub penalty assignment
• dummy node construction
• repeated randomized PCSF runs
• node and edge frequency aggregation
• network-supported metabolite annotation

Installation

You can install pcsf2featuremap from GitHub:

remotes::install_github("tidymass/pcsf2featuremap")

or:

pak::pak("tidymass/pcsf2featuremap")

Usage

This package is primarily intended to be used internally by featuremap.

The exported function is:

pcsf2featuremap::call_sr(
  from,
  to,
  cost,
  node_names,
  node_prizes
)

Arguments:

• from: character vector of edge source nodes
• to: character vector of edge target nodes
• cost: numeric vector of edge costs
• node_names: character vector of node names
• node_prizes: numeric vector of node prizes

The function returns the selected PCSF edges and nodes from the C++ solver.

Relationship to PCSF

The core solver in this package is adapted from the original PCSF R package:

https://github.com/IOR-Bioinformatics/PCSF

The original solver implements the prize-collecting Steiner forest heuristic described by Akhmedov et al.

This package is a lightweight extraction/wrapper of the core solver interface for use in featuremap. It does not reimplement the PCSF algorithm independently.

Citation

If you use this package or featuremap for PCSF-based analysis, please cite the original PCSF algorithm/software:

Akhmedov, M., Kedaigle, A., Chong, R. E., Montemanni, R., Bertoni, F., Fraenkel, E., & Kwee, I.
A fast prize-collecting Steiner forest algorithm for functional analyses in biological networks.

Please also cite featuremap where appropriate.

License

This package includes/adapts code from the original PCSF implementation. Please retain attribution to the original PCSF authors and repository when reusing or redistributing the solver.