The MetaBeeAI Project

About the Project

MetaBeeAI is an international consortium of bee ecotoxicologists and AI experts developing a pipeline that leverages large language models for automated systematic review and meta-analysis of research on bees and pesticides.

Our approach synthesizes the vast and heterogeneous body of published literature, encompassing studies on multiple bee species, pesticide types, experimental methodologies, and a wide range of toxicological endpoints. We are creating a novel data extraction and analysis pipeline that will provide an integrated perspective on the complex issue of pesticide impacts on pollinators.

Key Objectives

Develop automated tools for literature review and data extraction from published studies on bee-pesticide interactions

Understand how sublethal effects at different levels of biological organization impact overall fitness

Correlate sublethal findings with the lethal endpoint values used in environmental risk assessment

Create a framework that can be applied across the life sciences to integrate complex datasets

Improve understanding of the impacts of environmental stressors on pollinator populations

Methodology

The MetaBeeAI pipeline combines advanced natural language processing with domain expertise to:

Data Extraction

Using large language models to identify and extract relevant data from published literature, including experimental methodology, species information, pesticide exposure details, and measured endpoints.

Data Integration

Developing a standardized framework to compare findings across diverse studies with different experimental designs, endpoints, and reporting styles.

Analysis

Applying statistical techniques and machine learning to identify patterns, correlations, and potential causal relationships between pesticide exposure and various bee health endpoints.

Core Team