Computational oral biofilm pipeline developed during internship at NIFE (Niedersächsisches Institut für angewandte Zellgewebezüchtung), Hannover — part of the SIIRI/TRR-298 consortium.
North star: Hamilton/gLV ecological inference with metabolic sign priors → see PAPER_OUTLINE.md.
このリポジトリは、患者メタゲノム（ショットガンNGS）の菌叢データから、口腔バイオフィルムの代謝モデル（dFBA/COMETS）までを一気通貫でつなぐ計算パイプラインです。
NGS（shotgun）→ MetaPhlAn 4（菌叢プロファイル）→ init_comp.json（属レベル割合）
→ GEM（AGORAの代謝モデル）→ dFBA → COMETS（時間発展シミュレーション）
- まず試す: COMETS_beginner.ipynb（最低限の動作・可視化）
- パイプライン実行: run_comets_pipeline.py（Step A/B/C）
- 菌叢側（MetaPhlAn）: metaphlan_pipeline.sh → metaphlan_feature_table_to_init_comp.py
- 主な出力:
comets/pipeline_results/（図、COMETSの入出力、比較結果）
| 用語 | 意味（このrepoでの使い方） | 例（ファイル/コマンド） |
|---|---|---|
| NGS（shotgun） | メタゲノムのショットガンシーケンス。生データから菌叢組成を推定する入口 | data/ |
| MetaPhlAn 4 | リードから分類学的プロファイル（菌種/属の相対存在量）を推定 | qsub data/metaphlan_pipeline.sh |
| taxonomic profile | サンプルごとの菌叢組成（相対存在量の表） | MetaPhlAn出力 |
| init_comp.json | COMETS側の初期組成。対象7属の割合に正規化したJSON | data/metaphlan_feature_table_to_init_comp.py |
| GEM | Genome-scale metabolic model（ゲノム規模代謝モデル） | comets/agora_gems/ |
| AGORA | ヒト腸内細菌などのGEMコレクション。ここでは口腔細菌GEMを利用 | comets/agora_gems/*.xml |
| dFBA | 動的フラックスバランス解析。代謝（FBA）と環境（基質）の時間変化を結合 | comets/oral_biofilm.py |
| COMETS | 複数菌種の代謝・増殖を（空間あり/なしで）シミュレートする枠組み | comets/run_comets_pipeline.py |
| 0D / 2D | 0Dは空間なし（混合）。2Dは格子上で空間あり（拡散など） | Step A（0D）, Step B（2D） |
| cross-feeding | ある菌が作った代謝産物を別の菌が利用する現象 | 乳酸（lactate）など |
| Sobol感度解析 | パラメータの不確実性が結果へ与える寄与（全効果STなど）を推定 | qsub comets/run_sobol.sh |
| qsub / PBS | クラスタ投入（ジョブスケジューラ） | qsub ... |
This repository implements an end-to-end computational pipeline connecting patient metagenomic sequencing data to mechanistic biofilm simulation:
NGS (shotgun) → MetaPhlAn 4 → init_comp.json → GEM (AGORA) → dFBA → COMETS
The pipeline models multi-species oral biofilm on implant surfaces, focusing on the transition between commensal and dysbiotic states relevant to peri-implantitis.
pip install -e ".[dev]"
./scripts/reproduce_core.sh # LIGHT figures + ETL tests (no cluster)
make figures # matplotlib data figures only
pytest tests/ -q # metadata / QIIME2 ETL testsnife/
├── paper_data.py # Canonical paper-grade run paths (single source of truth)
├── guild_replicator_dieckow.py # 10-guild taxonomy + gLV ODE
├── scripts/
│ ├── fitting/ # gLV / Hamilton fits
│ ├── loo_cv/ # Leave-one-patient-out validation
│ ├── figures/ # Thesis + paper figure generators
│ ├── pde/ # Spatial diffusion / PINN
│ └── analysis/ # ML extensions (symbolic regression, attractor predictor)
├── dieckow_paper/ # Manuscript figures (make_figures.py)
├── comets/ # COMETS / dFBA simulation (5-species mechanistic)
├── jobs/ # PBS / GPU job scripts (cluster only)
├── tests/ # ETL unit tests (pytest)
├── docs/ARCHITECTURE.md # Data-flow diagram
├── docs/PROVENANCE.md # Per-figure regeneration commands
├── Makefile # `make figures` (LIGHT only)
└── data/ # NGS preprocessing scripts
| Code | Genus | Role |
|---|---|---|
| Str | Streptococcus spp. | Early colonizer, glucose→lactate |
| Act | Actinomyces / Schaalia | Scaffolding, early colonizer |
| Vel | Veillonella spp. | Obligate lactate cross-feeder (anaerobe) |
| Hae | Haemophilus parainfluenzae | Aerobic/facultative, NO₃ reducer |
| Rot | Rothia spp. | Health-associated, aerobic |
| Fus | Fusobacterium spp. | Bridge species, anaerobe |
| Por | Porphyromonas spp. | Late pathogen, deep anaerobe |
qsub data/metaphlan_pipeline.shRuns bowtie2 alignment against the AGORA/CHOCOPhlAn database, produces per-sample taxonomic profiles and converts them to init_comp.json (normalized genus fractions for the 7 target genera).
# Step A: 0D parameter sweep
python comets/run_comets_pipeline.py --step A
# Step B: 2D spatial healthy vs diseased comparison
python comets/run_comets_pipeline.py --step B
# Step C: Patient-specific (requires MetaPhlAn output)
python comets/run_comets_pipeline.py --step Cqsub comets/run_sobol.sh # Sobol indices, N=256, 12 paramsKey result: Fn_mu_max (ST=0.49) and Vp_Km_lac dominate dysbiosis — driven by lactate cross-feeding bridge, not Porphyromonas directly.
- Dieckow et al. 2024, npj Biofilms Microbiomes 10:155 — implant biofilm ground truth (volume, viability, composition)
- Dukovski et al. 2021, Nat. Protocols — COMETS framework
- Frings, Mukherjee et al. 2025, Analyst — ATR-FTIR strain-level identification of oral bacteria
- Joshi et al. 2025, npj Biofilms Microbiomes — peri-implantitis submucosal microbiome
| Doc | Purpose |
|---|---|
| docs/ARCHITECTURE.md | Core data flow + three pillars |
| docs/PROVENANCE.md | Figure regeneration (LIGHT vs HEAVY) |
| docs/ZENODO.md | Zenodo archive + citation |
| AGENTS.md | AI agent / slash-command entry points |
| CLAUDE.md | Full developer guide |
| PAPER_OUTLINE.md | Manuscript structure |
See CITATION.cff and docs/ZENODO.md. Interim BibTeX:
@software{nishioka2026nife,
author = {Nishioka, Keisuke},
title = {nife},
year = {2026},
url = {https://github.com/keisuke58/nife_intern},
version = {0.2.0}
}SIIRI / SFB TRR-298 — Safety Integrated and Infection Reactive Implants
Group: Prof. Meike Stiesch, MHH Department of Prosthetic Dentistry and Biomedical Materials Science
Experimental collaborators: Dr. Katharina Szafrański, Dr. Rumjhum Mukherjee, Dr. Pallavi Joshi