Market context: food packaging and human health protection
The food packaging sector is currently undergoing rapid regulatory, technological, and market-driven evolution.
Materials intended to come into contact with food (Food Contact Materials – FCM) must not only meet functional and sustainability requirements, but also ensure a high level of protection for human health, minimizing the risk associated with the migration of potentially hazardous chemical substances into food.
Many substances from food contact materials (FCM) can migrate into food, representing a potential source of chemical exposure for consumers. While some FCMs, such as plastics, are better characterized, most non-plastic FCM substances have not been fully evaluated for safety. In addition, migration of non-intentionally added substances (NIAS), including impurities, degradation products, and neoformed compounds, occurs in both plastic and non-plastic FCMs. These NIAS often lack toxicological characterization, making their safety assessment a challenge. In this context, predictive approaches based on chemical structure, such as (Q)SAR models, are increasingly used to estimate toxicological properties like genotoxicity or carcinogenicity in a fast, cost-effective, and animal-free way.
Although tools like EFSA’s RACE methodology can support risk evaluation of FCM-related contaminants, data collection is time-consuming, and the approach does not specifically address the endocrine-disrupting potential of substances. To overcome these limitations, the SILIFOOD tool was developed, integrating existing regulatory and toxicological information with in-silico predictions to enable a more streamlined and rapid risk assessment of non-evaluated FCM substances. Its applicability was demonstrated through four case studies on non-evaluated substances migrating from FCMs, including non-plastic FCM compounds and NIAS.
The SILIFOOD project and the contribution of Kode Chemoinformatics
The SILIFOOD project was conceived to develop an innovative in-silico tool to support the risk assessment of substances not yet evaluated and present in materials intended to come into contact with food. The system integrates regulatory data, existing toxicological information, and predictive models based on (Quantitative) Structure–Activity Relationships ((Q)SAR), enabling semi-automated collection and analysis of relevant data.
Within the project, Kode Chemoinformatics, the Business Unit of Kode, contributed to the development of the chemoinformatics tool, providing advanced expertise in chemical database management, predictive modeling, and the integration of computational workflows to support chemical safety assessment.
The project was developed in collaboration with Sciensano and the Mario Negri Institute for Pharmacological Research IRCCS, representing a concrete example of how chemoinformatics can be effectively applied to real-world challenges in the food packaging sector.
Practical value for food packaging manufacturers and R&D departments
The SILIFOOD tool enables the interrogation and integration of dozens of chemical and toxicological databases, covering thousands of organic substances potentially present in food contact materials. Through (Q)SAR models and read-across strategies, the system provides predictive evaluations for key toxicological endpoints, including:
- bioavailability and bioaccumulation potential
- genotoxicity
- carcinogenicity
- developmental and reproductive toxicity
- sub-chronic toxicity
- endocrine activity
- Cramer class identification
The outcome is the generation of structured and traceable reports, supporting substance screening, new material design, and the prioritization of experimental testing. This approach represents a tangible competitive advantage for food packaging manufacturers and R&D departments, reducing time, costs, and uncertainty in chemical safety evaluation processes.
Scientific publication
The results of the SILIFOOD project and the development of the tool have been described in a scientific article published on Science Direct, confirming the robustness of the methodology and the scientific value of the proposed approach.
Read the paper: https://www.sciencedirect.com/science/article/pii/S3050620425000673?via%3Dihub