GeneCover

A combinatorial approach for label-free marker gene selection from scRNA-seq and spatial transcriptomics data, using gene-gene correlation and set cover.

This repository is a Python packaging + PyPI-friendly version of the original GeneCover pipeline, keeping the algorithm behavior as close as possible to the original implementation while supporting optional solver backends (uses lazy loading).

---

Manuscript

Wang, A., Hicks, S., Geman, D., & Younes, L. (2025, April).
GeneCover: A Combinatorial Approach for Label-free Marker Gene Selection.
In International Conference on Research in Computational Molecular Biology (pp. 354–357). Cham: Springer Nature Switzerland.
DOI: 10.1101/2024.10.30.621151

@inproceedings{wang2025genecover,
  title={GeneCover: A Combinatorial Approach for Label-free Marker Gene Selection},
  author={Wang, An and Hicks, Stephanie and Geman, Donald and Younes, Laurent},
  booktitle={International Conference on Research in Computational Molecular Biology},
  pages={354--357},
  year={2025},
  organization={Springer}
}

Original resources:

• Original GitHub repo: https://github.com/ANWANGJHU/GeneCover
• Original docs site: https://genecover.readthedocs.io/

---

This package exists to provide a lightweight, pip-installable interface to GeneCover without requiring users to clone the original repository or configure solvers unless needed.

Features

• Compute gene–gene correlation matrices from scRNA-seq / spatial transcriptomics data:
  • Spearman correlation (default)
  • Pearson correlation
• Select marker genes via minimal-weight set cover, with multiple backends:
  • Gurobi (integer programming; requires license)
  • SCIP via PySCIPOpt (open-source solver)
  • Greedy heuristic (no solver dependencies; NumPy-only)
• Perform iterative marker selection (Iterative_GeneCover) to build gene panels in multiple rounds
• Behavior closely aligned with the original one-file GeneCover implementation

---

Compatibility with the original GeneCover implementation

• The core GeneCover algorithm, thresholds, and optimization logic are preserved.
• gene_gene_correlation, GeneCover, and Iterative_GeneCover follow the same behavior and defaults as the original implementation.
• Differences from the original repository are limited to packaging and dependency handling (lazy imports, optional solver extras).
• Results should be directly comparable to those obtained using the original GeneCover code.

---

Installation

Install from PyPI

pip install genecover

This installs the core package and supports the Greedy backend (no external solver required).

Optional solver backends

To use integer-programming solvers, install extras:

Gurobi backend (requires a valid Gurobi license)

pip install "genecover[gurobi]"

SCIP backend (via PySCIPOpt)

pip install "genecover[scip]"

Install all optional backends

pip install "genecover[all]"

---

Usage and QuickStart

For full tutorials and usage examples, see the original GeneCover documentation. The function names should be the same, for instance:

from genecover import gene_gene_correlation, GeneCover, Iterative_GeneCover

---

Project Status

This package focuses on: a clean Python API (genecover), reproducibility with the original one-file pipeline, and solver backends as optional dependencies. The interface may evolve as packaging/testing/docs are improved.

---

Contributing

Bug reports, feature requests, and GitHub issues or pull requests are welcome.

Please submit issues and pull requests via the GitHub repository: https://github.com/danielchen05/GeneCover

---

Note

This project has been set up using PyScaffold 4.6. For details and usage information on PyScaffold see https://pyscaffold.org/.