MarkerFinder

Adaptive HGT-Aware Phylogenomic Pipeline for Prokaryotic Marker Gene Selection

English README | 中文说明 | Manual (EN) | Manual (CN)

Documentation in this repository ships in both languages, with the English version as the primary text. tests/unit/test_docs_are_bilingual.py fails if a document exists in only one language, if an "English" file is written in Chinese, or if the two versions diverge in shape.

Overview

MarkerFinder is a phylogenomic pipeline for prokaryotic marker gene selection and species tree inference. It targets three common challenges in existing methods:

1. Rigid marker sets — MarkerFinder dynamically selects optimal markers adapted to your dataset
2. HGT contamination — phylogenetic-only horizontal gene transfer screening removes phylogenetically misleading markers
3. MAG fragility — quality-aware adaptive parameters handle incomplete metagenome-assembled genomes

Conceptual Origin

MarkerFinder's integrative (concatenation + coalescence) marker-consistency idea is methodologically derived from Steenwyk & King (2025), Science 390, 751–756, doi:10.1126/science.adw9456 — a paper retracted at the authors' own request on 2026-02-05 (retraction notice: Science 391, 564, doi:10.1126/science.aef5589). The retraction concerns the paper's biological conclusion, not its methodological framework; MarkerFinder borrows only the latter and implements, evaluates and defends it independently. See README.EN.md § Conceptual Origin (中文) for the full citation and the element-by-element mapping into this codebase.

Quick Start

# Install (into an environment that also has the external tools)
conda env create -f environment.yml
pip install -e ".[dev]"

# Run
markerfinder \
    -i genomes/ \
    -o output/ \
    -t 8 \
    --mode standard \
    --gtdb-markers-dir gtdbtk_out/align/marker_genes

Documentation

Testing

The repository carries three test layers. The first two are fast and self-contained; the third runs the real pipeline on real data and is the acceptance evidence published with the release.

validation/capabilities.py + validation/cases/test_v90_capability_coverage.py make coverage checkable: every option, subcommand, exit code and output product the software advertises is enumerated from the code, and any entry without a case fails the suite. The exported matrix lives in validation/results/capability_matrix.tsv.

Author

Zengzichao (曾子超) — zengzichao@sjtu.edu.cn · ORCID 0000-0001-6553-970X

License

MIT