khaoulabz/InSilicoChemoTax: InSilicoChemoTax v1.0.0

🔬 InSilicoChemoTax v1.0.0 — Initial Public Release

A multi-method machine learning framework for genome-based chemotaxonomic profiling in polyphasic bacterial taxonomy

What's Included

• InSilicoChemoTax.ipynb — Full pipeline (15 code cells, 31 total cells)
• environment.yml — Conda environment for reproducibility
• data/template_input.xlsx — Example input file (EC counts, BV-BRC format)
• README.md — Full documentation and quick start guide

Pipeline Capabilities

• 5 independent analytical approaches — Euclidean distance, Mahalanobis distance (Ledoit-Wolf regularization), hierarchical clustering (Ward's method), UMAP 3D/2D, Random Forest classification
• Full statistical validation — PERMANOVA, Betadisper, LOO-CV, bootstrap resampling (n=500), leave-one-genus-out (LOGO)
• Auto-generated outputs — IJSEM-compliant Methods + Results text, publication-quality figures (4 main + 5 diagnostic), supplementary tables (3 CSV files)

Validated On

• 124 actinobacterial type strains across 5 genera
• 107 informative EC numbers from 9 KEGG chemotaxonomic pathways
• Novel strain: Streptomyces sp. Mg15
• Consensus accuracy: 5/5 methods unanimous

Environment

| Package | Version | |---|---| | Python | 3.10.19 | | scikit-learn | 1.7.2 | | NumPy | 2.2.6 | | SciPy | 1.15.2 | | pandas | 2.3.3 | | umap-learn | 0.5.11 |

Citation

If you use this release, please cite:

Bouznada, K. (2026). InSilicoChemoTax: A multi-method machine learning framework for genome-based chemotaxonomic profiling (v1.0.0). Zenodo. doi.org

Known Limitations

• Validated on actinobacterial type strains — generalizability to other phyla requires independent validation
• Input requires manual EC count retrieval from BV-BRC Pathways viewer
• Pathway set (9 KEGG pathways) optimized for actinobacterial chemotaxonomy — may require adaptation for other groups

Contact

Dr. Khaoula Bouznada — LBSM Laboratory, Algiers, Algeria
📧 [khaoula.bouznada@g.ens-kouba.dz]