Warmr: AI in Ecology and Epidemiology

Project overview

This project focuses on utilising the capabilities of Large Language Models (LLMs) to build intuitive, conversational tools for analyzing ecological and epidemiological datasets.

Using Lassa fever data provided by the Nigerian Centre for Disease Control as a case study, we aim to automate complex data analysis processes and make them accessible to non-expert users, such as policymakers and public health officials.

We demonstrate how novel technology can improve decision-making in ecological and public health contexts by developing a Virtual Data Scientist, a chatbot powered by advanced AI workflows.

Our aims

• To develop an interactive AI-powered tool that leverages Large Language Models (LLMs) to enhance access to ecological and epidemiological datasets.
• To create a Virtual Data Scientist capable of supporting non-technical users, such as policymakers, by automating data querying, visualization, and analysis.
• To integrate and refine multi-agent workflows for handling diverse tasks, including database queries, statistical computations, document retrieval and automated report writing based on requested data.
• To demonstrate the practical application of these methods using Lassa fever as a real-world case study, showcasing the tool’s potential for broader ecological and public health challenges.

Who is involved?

This project, funded by Wellcome, is a collaboration primarily with the University of Leicester.

Research assistant Artur Trebski is responsible for automating epidemiological reporting in our lab, using data provided by partners such as the Nigerian Centre for Disease Control.

Our methods

Multi-Agent System Design

The backbone of this project is a multi-agent system built using LangChain and LangGraph frameworks to streamline the use of Large Language Models (LLMs) in automating complex tasks.

This system employs agents, which are specialised modules designed for specific functions. These agents work collaboratively to process and analyse ecological and epidemiological data, with Lassa fever serving as a case study.

Key agents in our workflow include:

• Primary Assistant: Interprets user queries and routes them to the appropriate agent.
• SQL Agent: Retrieves structured data from databases.
• Coding Agent: Conducts data analysis and visualisation.
• Knowledge Agent: Extracts relevant insights from indexed scientific literature.

The system processes epidemiological data, such as Lassa fever cases, travel times, and environmental covariates, alongside contextual knowledge from the scientific literature to deliver holistic insights.

Designed as a conversational chatbot interface, it allows users to interact naturally. They can ask queries such as “Show me annual lassa fever cases in three regions,” which are efficiently handled through coordinated agent workflows.

This approach combines AI with accessible design, reducing the technical complexity of data analysis and making advanced tools available to researchers, policymakers and other stakeholders.

Our role

The Museum’s role in the project includes:

• Designing and implementing a multi-agent system for automated data analysis using LangGraph and LangChain frameworks.
• Integrating epidemiological data and contextual knowledge to support decision-making in ecology and public health.
• Collaborating with the Nigerian Centre for Disease Control to validate the system using Lassa fever case data.
• Demonstrating the application of AI-driven tools for ecological and epidemiological research.
• Communicating the development and potential of the tool through conferences, participation in AI networks and collaborations.