[On digitalizing the analysis of food chemistry composition data]
E A Smirnova1, V V Bessonov1, E S-A Shakhvalieva1
1Federal Research Centre for Nutrition, Biotechnology and Food Safety, 109240, Moscow, Russian Federation.
Abstract:
The development of digital nutrition science, based on the application of artificial intelligence algorithms for the personalization of dietary recommendations, is hindered in the Russian Federation by the absence of a unified verified database of the chemical composition of foods. Such a resource should representatively reflect the current state of the agro-industrial complex and include products that form the basis of the diet of various demographic groups of the population. The objective of the research was to develop a software-analytical complex for creating and verifying a national database of the chemical composition of food. The system being created will ensure the consolidation of data from heterogeneous sources on food raw materials, industrially produced products, and culinary dishes, forming a representative resource that reflects the structure of the actual nutrition of the population and the current product range.
Material And Methods:
The development of a three-tier software-analytical complex (MySQL 8.0, Python 3.13.3/Flask 2.0, HTML5/CSS3/JavaScript) involved the implementation and refinement of the following approaches: data collection (via web interface and REST API), data verification (using algorithms for missing value imputation, outlier detection, and k-means clustering), and the creation of algorithms to calculate the nutritional value of foods and dishes, accounting for technological losses. The system was validated against retrospective data from nutritional epidemiology studies.
Results:
A software-analytical complex has been created, including three interconnected databases: the chemical composition of food raw materials and industrially produced foods, dishes and culinary products, as well as coefficients of losses during technological processing. Data verification algorithms were developed and implemented, including methods for processing missing values, detecting statistical outliers, and clustering. A module for calculating the nutritional value of ready-made dishes, taking into account technological losses, was created. Criteria for classifying data as verified were developed, including requirements for completeness of filling (≥95%), compliance with permissible value ranges, and consistency with reference values. Tools for calculating the nutritional value of ready-made dishes, taking into account technological losses, were created, including a methodology for selecting recipes considering regional dietary characteristics.
Conclusion:
The developed software-analytical complex with a web interface is designed for managing data on the chemical composition of food. The system supports the full data lifecycle. Further development of the complex involves expanding the list of analyzed nutrients and integration with software for various purposes, including for dietitians, public catering specialists, diet calculations for organized groups, as well as applications for individual nutrition assessment and informing the population about the energy and nutritional value of diets. The implementation of the project will provide an evidence base for epidemiological research, the development of preventive measures for diet-related diseases, and the formation of a scientifically based state policy in the field of healthy nutrition.


