humaninfinder 1.0.5

A professional evolutionary discovery tool for Humanin-like peptides (sORFs) using a Hybrid AI approach.

HumaninFinder v1.0.5 🧬🤖

---

HumaninFinder is a professional, high-performance Python framework designed for the discovery and classification of Humanin-like peptides (sORFs) within mitochondrial genomes. It employs a Hybrid AI Engine that integrates deep structural embeddings from the ESM-2 Protein Language Model with explicit biophysical analysis to identify functional, non-canonical, and pseudogenic sequences across any taxonomic group.

---

📂 Repository Structure

.
├── conda/                   # Bioconda recipe and metadata
├── deploy/                  # Containerization (Dockerfile, Singularity.def)
├── docs/                    # Technical documentation and user guides
├── examples/                # Quick-start samples (FASTA genomes)
├── galaxy/                  # Galaxy Tool wrapper and integration
├── paper/                   # Publication manuscripts
│   ├── joss/                # Software description for JOSS
│   └── primate_study/       # Scientific case study on 61 primate genomes
├── src/
│   └── humaninfinder/       # Main Python Package
│       ├── cli.py           # Subcommand-based Command-line interface
│       ├── core.py          # Locus localization and ORF finding logic
│       ├── classifier.py    # Hybrid AI Engine (ESM-2 + Biophysical)
│       ├── agent.py         # AI Research Agent (Ollama integration)
│       ├── data/            # HMM models and 16S probes
│       └── models/          # Pre-trained hybrid classifier weights
├── tests/                   # Unit and biological validation tests
├── pyproject.toml           # Build system and PyPI definitions
└── pixi.toml                # Modern environment management

🧩 Core Components Detail

• Hybrid AI Engine: Combines mean-pooled embeddings from the ESM-2 transformer model (esm2_t6_8M_UR50D) with charge, pI, and hydrophobicity metrics.
• Evolutionary Rescue: A high-sensitivity sliding-window scanner that "rescues" non-canonical and pseudogenic relics in diverged lineages.
• AI Research Agent: An integrated specialist assistant powered by local LLMs (via Ollama) to provide biological interpretation of results in the context of mitochondrial aging and cytoprotection.
• Biological Deduplication: A specialized filter that ensures independent evolutionary signals by removing technical windowing artifacts.

---

🚀 Key Features

• Organism Agnostic: Supports all 33 NCBI genetic codes, enabling MDP discovery in any mitochondria-bearing taxon.
• Expert AI Agent: Built-in specialist in Humanin, mitochondrial signaling, and aging biology to interpret your findings.
• High-Throughput Ready: Parallelized processing and optimized inference for large genomic collections.
• Validated Science: Built-in reproduction of the 61-primate evolutionary case study.

---

🛠️ Quick Start

1. Installation

Option A: via pip (Fastest)

pip install "humaninfinder[agent]"
humanin-finder setup

Note: Ensure HMMER3 is installed on your system.

Option B: via Conda / Mamba (Recommended)

Perfect for an isolated scientific environment:

# Create environment from the provided file
mamba env create -f environment.yml
mamba activate humanin_env

# Finalize setup
humanin-finder setup

2. Run Discovery Pipeline

humanin-finder predict -i genome.fasta -o results --hmm --rescue

3. Biological Interpretation

# Get a summary of your findings from the AI Specialist
humanin-finder agent --results results_results.csv

---

📖 Documentation

Detailed information is available in the paper/ directory:

• 🛠️ Software Paper: Architecture and methodology for JOSS.
• 🏗️ Scientific Report: Evolutionary dynamics of Humanin in Primates.

---

🤝 Contributing

Contributions are welcome! Please see our CONTRIBUTING.md for details.

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

---

Developed by LaBiOmicS - Laboratory of Bioinformatics and Omics Sciences. Institution: Universidade de Mogi das Cruzes (UMC)