AlgicideDB: a comprehensive database enhanced by large language models for algicide management and discovery.

Area of Science:

• Environmental Science
• Ecotoxicology
• Computational Biology

Background:

• Harmful algal blooms (HABs) present escalating global threats to aquatic ecosystems and human health.
• Chemical algicides offer rapid control but face limitations due to data scarcity and environmental toxicity concerns.

Purpose of the Study:

• To develop AlgicideDB, a comprehensive database of algicides and their properties.
• To create tools for exploring algicidal compounds and assessing their ecological risks.
• To integrate an AI-powered system for HAB-related query resolution and knowledge dissemination.

Main Methods:

• Manual curation of 1,672 algicidal records on 542 algicides targeting 110 algal species.
• Analysis of algicide physicochemical properties and development of an algicide-likeness scoring function.
• Evaluation of acute toxicity of algicides to non-target aquatic organisms across trophic levels.
• Implementation of a large language model (LLM) with retrieval-augmented generation (RAG) for HAB queries.

Main Results:

• AlgicideDB provides a systematic resource for algicide data, including physicochemical properties and toxicity information.
• A novel scoring function aids in identifying potential antialgal compounds.
• Ecological risk assessment data for various algicides is available.
• The integrated LLM-RAG system offers an advanced tool for HAB information retrieval and decision support.

Conclusions:

• AlgicideDB is a valuable, innovative platform for exploring algicidal compounds.
• The database facilitates the development of safer and more effective strategies for managing harmful algal blooms.
• AlgicideDB supports informed decision-making and knowledge dissemination in HAB control.