[Paper Review] A Platform for the Biomedical Application of Large Language Models

Sebastian Lobentanzer, Feng, Shaohong+7 more|arXiv (Cornell University)|May 10, 2023

Artificial Intelligence in Healthcare and Education13 citations

TL;DR

The paper presents BioChatter, an open-source framework to interface with biomedical LLMs, integrating retrieval-augmented generation, model chaining, benchmarking, and privacy-preserving deployment, demonstrated via two web apps.

ABSTRACT

Current-generation Large Language Models (LLMs) have stirred enormous interest in recent months, yielding great potential for accessibility and automation, while simultaneously posing significant challenges and risk of misuse. To facilitate interfacing with LLMs in the biomedical space, while at the same time safeguarding their functionalities through sensible constraints, we propose a dedicated, open-source framework: BioChatter. Based on open-source software packages, we synergise the many functionalities that are currently developing around LLMs, such as knowledge integration / retrieval-augmented generation, model chaining, and benchmarking, resulting in an easy-to-use and inclusive framework for application in many use cases of biomedicine. We focus on robust and user-friendly implementation, including ways to deploy privacy-preserving local open-source LLMs. We demonstrate use cases via two multi-purpose web apps (https://chat.biocypher.org), and provide documentation, support, and an open community.

Motivation & Objective

• Motivate the need for safe, accessible interfacing with large language models in biomedicine.
• Propose an open-source framework that unifies current LLM capabilities (RAG, chaining, benchmarking) for biomedical use cases.
• Ensure robust, user-friendly deployment including privacy-preserving local LLM options.
• Provide documentation, support, and an open community to foster adoption and collaboration.

Proposed method

• Leverages open-source software packages to build an interoperable framework.
• Integrates knowledge integration / retrieval-augmented generation with model chaining and benchmarking.
• Emphasizes privacy-preserving deployment of local open-source LLMs.
• Demonstrates applicability through two multi-purpose web applications and accompanying documentation.
• Aims for an inclusive, easy-to-use interface suitable for diverse biomedical use cases.

Experimental results

Research questions

• RQ1 How can an open-source platform effectively integrate retrieval-augmented generation, model chaining, and benchmarking for biomedical LLM applications?
• RQ2 What design choices enable robust, privacy-preserving deployment of local open-source LLMs in biomedicine?
• RQ3 How can bioscience researchers leverage web apps and documentation to adopt BioChatter for varied biomedical tasks?

Key findings

• BioChatter demonstrates use cases via two multi-purpose web applications (accessible via a provided URL).
• The platform provides documentation, support, and an open community to facilitate adoption.
• The work focuses on robust, user-friendly implementation and the integration of multiple LLM functionalities for biomedical use.