gaia-soil 0.1.0

A foundation model for soil microbiome understanding

Gaia: Soil Microbiome Foundation Model

Gaia — the Greek goddess of Earth. Decoding the hidden language of soil microbiomes.

"The AlphaFold of Soil Microbiomes, built open-source."

English | 한국어

Gaia is a foundation model that understands the "language" of soil microbial communities. Pre-trained on public metagenomic data, it enables soil health diagnosis, yield prediction, and microbial consortium design.

---

Key Features

• Pre-trained Foundation Model: Transformer-based model pre-trained on 10,000+ soil microbiome samples from MGnify, NEON, and EMP
• Soil Health Diagnosis: Predict soil chemical properties (pH, organic carbon, total nitrogen) from microbial profiles
• Biome Classification: Identify soil biome types (agricultural, forest, grassland, desert, wetland)
• Drought Stress Detection: Binary classification of drought stress from microbial signatures
• Interpretability Tools: Attention-based keystone genera identification
• Synthetic Data Generation: Generate realistic microbial abundance profiles for target soil conditions

Quick Start

Installation

pip install gaia-soil

Or install from source:

git clone https://github.com/Kimchikilla/ProjectGaia.git
cd ProjectGaia
pip install -e ".[dev]"

Basic Usage

from gaia.inference import GaiaPredictor

# Load pre-trained model
predictor = GaiaPredictor.from_pretrained("gaia-v0.1")

# Predict soil properties from microbial profile
result = predictor.diagnose("path/to/abundance_profile.csv")
print(result.soil_health_report)

Project Structure

gaia/
├── README.md
├── LICENSE                    # Apache 2.0
├── CONTRIBUTING.md
├── docs/
│   ├── roadmap.md
│   ├── data_standard.md       # Data standardization guide
│   └── tutorials/
├── data/
│   ├── scripts/               # Data collection & preprocessing scripts
│   ├── configs/               # Data source configurations
│   └── README.md              # Data catalog
├── gaia/
│   ├── preprocessing/         # Preprocessing modules
│   ├── models/                # Model architectures
│   ├── training/              # Training scripts
│   ├── evaluation/            # Evaluation modules
│   └── inference/             # Inference modules
├── benchmarks/                # Benchmark datasets & evaluation criteria
├── notebooks/                 # Tutorial Jupyter notebooks
└── tests/

Data Sources

Source	Description	Samples
MGnify	Taxonomic abundance tables from soil biomes	5,000-15,000
NEON	Paired microbiome + environmental data	~2,000
Earth Microbiome Project	Standardized global soil samples	~5,000
SMAG	40,039 soil MAGs from 3,304 metagenomes	Reference DB

Benchmarks

Task	Metric	Description
Biome Classification	ROC-AUC, F1	Classify soil biome type from microbial profile
Soil Chemistry Prediction	R², RMSE	Predict pH, organic C, total N
Tillage Classification	Accuracy, Kappa	Classify tillage practice
Drought Stress Detection	Accuracy, F1	Detect drought stress (binary)
Abundance Reconstruction	Cosine Similarity	Reconstruct masked microbial profiles

Model Architecture

• Base: Multi-layer Transformer Decoder
• Layers: 6-12 (adjustable)
• Attention Heads: 8-16
• Embedding Dim: 256-512
• Vocabulary: ~5,000 soil-associated genera
• Pre-training: Continual pre-training from MGM weights

Tech Stack

Area	Tool
Language	Python 3.10+
Deep Learning	PyTorch 2.x
Transformers	Hugging Face Transformers
Data	Pandas, AnnData, Biom-format
Bioinformatics	QIIME2, Kraken2, MetaPhlAn
Visualization	Matplotlib, Seaborn, UMAP
Experiment Tracking	Weights & Biases
Model Hosting	Hugging Face Hub

Contributing

We welcome contributions! Please see CONTRIBUTING.md for guidelines.

Ways to Contribute

• Code: Bug fixes, new features, pipeline improvements
• Data: Standardized soil microbiome datasets
• Science: New benchmark tasks, ecological validation, domain expertise

Community

• GitHub Discussions: Technical discussions and Q&A
• Discord: Real-time community chat
• Monthly Meetings: Online direction-setting meetings (1st Thursday of each month)

Citation

@software{gaia2026,
  title={Gaia: A Foundation Model for Soil Microbiome Understanding},
  year={2026},
  url={https://github.com/Kimchikilla/ProjectGaia}
}

License

This project is licensed under the Apache License 2.0 - see LICENSE for details.

---

This project is under active development. Star this repo to stay updated!