bsllmner-mk2

A tool for extracting biological named entities from BioSample records using Large Language Models (LLMs) and mapping them to ontology terms.

Key capabilities:

• Extract mode -- Performs Named Entity Recognition (NER) to extract terms such as cell line, tissue, and organism from BioSample metadata
• Select mode -- Extends extract mode by mapping extracted terms to ontology entries (Cellosaurus, UBERON, Cell Ontology, etc.)

bsllmner-mk2 uses Ollama as the LLM inference server.

Quick Start

docker compose up -d --build
docker compose exec app bsllmner2_extract \
  --bs-entries tests/data/example_biosample.json \
  --model llama3.1:70b --debug

For a complete walkthrough including ontology setup and Select mode, see Getting Started.

Documentation

Full documentation is available at https://dbcls.github.io/bsllmner-mk2.

Basics

• Getting Started -- First run walkthrough with ontology setup
• Installation -- Docker Compose, uv, and GPU configuration

Features

• Extract Mode -- NER extraction pipeline and CLI options
• Select Mode -- Ontology mapping pipeline and CLI options
• Data Formats -- Input/output data format specification
• Configuration -- Environment variables and settings

Operations

• ChIP-Atlas -- Processing ChIP-Atlas data (hg38/mm10)
• NIG Slurm -- Running on NIG Slurm environment

Development

• Development -- Development environment setup
• Testing -- Unit tests, linting, mutation testing, model evaluation

Related Resources

• Original repository: sh-ikeda/bsllmner
• Related paper: https://doi.org/10.1101/2025.02.17.638570

Other Interfaces

bsllmner-mk2 also includes a FastAPI server (bsllmner2_api) and a React-based web UI, but these are not actively maintained and their operation is unverified.

License

This repository is released under the MIT License. For details, see the LICENSE file.