Scientific Documentation
The Pathogens in Foods (PIF) Database is an open-access scientific resource designed to support food safety surveillance, evidence synthesis, and quantitative microbial risk assessment. The publications below describe the scientific foundations of the database, including its FAIR-compliant data architecture, systematic review methodology, harmonized data model, artificial intelligence capabilities, and strategic roadmap for future development.
Together, these publications provide a comprehensive reference for researchers, food safety authorities, risk assessors, and other stakeholders interested in understanding how PIF is developed, maintained, and applied to support evidence-based decision making in food safety.
Pathogens-in-Foods Database: Training Toolkit
This European Food Safety Authority (EFSA) publication presents the training materials developed to support the uptake and practical use of the PIF Database in food safety risk assessment. The toolkit introduces the concept, structure, and FAIR principles of PIF and includes practical exercises requiring users to access and analyse PIF data to address microbiological food safety questions. The materials are intended to support dissemination and capacity-building activities by EFSA focal points, programmes such as EU-FORA, national food safety agencies, and other stakeholders, helping increase user engagement and promote the use of harmonised, systematically reviewed occurrence data in food safety research and risk assessment.
Publisher: European Food Safety Authority (EFSA)
A Retrieval-Augmented Natural Language Interface for Data Description and Meta-Analysis in the Pathogens-in-Foods (PIF) Database
This article presents the next generation of the PIF platform by integrating large language models with retrieval-augmented generation (RAG) and deterministic statistical analysis. The proposed interface enables users to query the database using natural language, automatically generate evidence summaries, and perform reproducible meta-analyses without requiring programming expertise. The study evaluates several open-source and proprietary language models, demonstrating reliable tool selection, accurate data retrieval, and grounded AI-assisted analytical workflows for food safety research.
Journal: Journal of Food Protection
Feasibility Study on the Pathogens-in-Foods Database
This European Food Safety Authority (EFSA) scientific report evaluates the long-term sustainability, adoption potential, and future evolution of the PIF Database. The study assesses technical, organizational, legal, and ethical aspects of the platform and proposes a strategic roadmap that includes the expansion of pathogen catalogues, artificial intelligence for data analytics, antimicrobial resistance (AMR) data integration, and the application of large language models to support systematic reviews and data verification.
Publisher: European Food Safety Authority (EFSA)
Development of an Intelligent Agent for Knowledge Extraction in the Pathogens in Foods (PIF) Database with Machine Learning
This Master's dissertation presents the design and development of a Visual Natural Language Interface (V-NLI) for the PIF Database, enabling food safety experts to explore complex microbiological occurrence data without requiring programming expertise. The resulting PIF Intelligent Agent combines tool-calling Small Language Models (SLMs), Retrieval-Augmented Generation (RAG), vector search, and prompt engineering to support natural-language querying, meta-analysis, data visualization, and automated scientific report generation. The system adopts a hybrid architecture that separates language interpretation from deterministic statistical computation, with an Open Chat Mode for flexible exploratory analysis and a Guided Meta-Analysis Mode for structured, reproducible scientific workflows. The dissertation also evaluates multiple language models, demonstrating reliable tool execution across different model sizes and highlighting interpretive fidelity, factual coherence, and conciseness as important factors in AI-assisted scientific analysis.
Degree: Master's Degree in Informatics
Pathogens-in-Foods (PIF): An Open-Access European Database of Occurrence Data of Biological Hazards in Foods
This publication introduces the scientific foundations of the PIF Database. It describes the complete data lifecycle, from protocol-driven systematic reviews and standardized data extraction to quality assurance and publication under the FAIR principles (Findability, Accessibility, Interoperability, and Reusability). The paper also presents the PIF system architecture, harmonized data model, controlled vocabularies, and the novel CCC data quality framework (Consistency, Conformity, and Completeness) developed to ensure reliable and reusable occurrence data for food safety research.
Journal: Microbial Risk Analysis
