COMBATTING DIET RELATED NON-COMMUNICABLE DISEASE THROUGH ENHANCED SURVEILLANCE

Project Information

CoDiet

Grant agreement ID: 101084642

DOI

10.3030/101084642

EC signature date 17 October 2022

Start date 1 January 2023

End date 31 December 2026

Funded under

• Food, Bioeconomy Natural Resources, Agriculture and Environment

Total cost

€ 7 330 589,25

EU contribution

€ 7 330 589,25

Coordinated by

FUNDACION AZTI - AZTI FUNDAZIOA
Spain

Periodic Reporting for period 2 - CoDiet (COMBATTING DIET RELATED NON-COMMUNICABLE DISEASE THROUGH ENHANCED SURVEILLANCE)

Reporting period: 2024-07-01 to 2025-12-31

Summary of the context and overall objectives of the project

Unhealthy diets drive metabolic disruption and raise non-communicable disease (NCD) risk, yet the mechanisms are poorly defined and current dietary assessment tools are imprecise, with sparse data for vulnerable groups. The EU-funded CoDiet project addresses these gaps by building AI-enabled dietary assessment and a population-level diet–NCD monitoring tool to support uptake of NCD-protective diets.
CoDiet will deliver tools across eight work packages to: clarify diet–NCD mechanisms; improve dietary measurement; integrate multi-omics and sensor data; and translate evidence into policy-ready outputs. Six objectives guide the work:
1: AI-driven literature discovery to map physiological and metabolic links between diet and NCD.
2: Deepen understanding of NCD risk factors through beyond-state-of-the-art mechanistic methods.
3: Quantify individual variation in dietary responses to inform targeted advice.
4: Develop enhanced dietary assessment via machine learning.
5: Build an improved diet–NCD monitoring tool for population surveillance.
6: Create a dynamic interface between monitoring and policy to ensure population impact.
These activities open a pathway to increase adoption of NCD-protective diets at scale.

Work performed from the beginning of the project to the end of the period covered by the report and main results achieved so far

CoDiet has progressed from foundational data generation to large‑scale integration, advanced modelling and clinical readiness. A major achievement has been the creation of the largest fully double‑annotated biomedical text corpus in nutrition and NCD research, enabling state‑of‑the‑art relation‑extraction workflows based on deployable open‑weight language models. This work underpins a transparent knowledge graph integrating more than 4,000 scientific publications to support causal inference and interpretation of diet–disease mechanisms.

In parallel, data from the multi‑country pilot study in Spain, the United Kingdom, Ireland and Greece have been fully cleaned, harmonised and curated. These datasets include dietary records, metabolomics, microbiome profiles, lipidomics, non‑invasive sensor data, passive‑camera images and clinical metadata. Secure sharing through the project’s CESNET data infrastructure has enabled integrated statistical analyses, machine‑learning development and the construction of a personalised dietary recommendation system.

Substantial advances have been made in analytical and computational infrastructure, including distributed and federated learning, efficient execution of large‑scale pipelines and automation of multimodal data alignment. Integrated modelling of dietary intake, multi‑omics data and cardiometabolic indicators has identified biologically meaningful associations relevant to NCD risk, supporting participant stratification and personalised nutrition approaches.

The personalised recommendation system has reached an advanced stage, combining a participant‑facing mobile application, a data‑driven recommendation engine, behavioural messaging and adaptive follow‑up to support adherence. Recruitment for the multi‑country intervention study has begun under a harmonised protocol, with participants enrolled and biological and imaging samples being processed in preparation for the intervention phase. Continuous ethical oversight and strong scientific coordination have ensured responsible project progress.

Progress beyond the state of the art and expected potential impact (including the socio-economic impact and the wider societal implications of the project so far)

CoDiet delivers advances that go beyond current practice in dietary assessment, multi‑omics integration, AI‑driven modelling and policy simulation by combining objective measurement, mechanistic insight and scalable digital tools within a single, coherent framework.

• Knowledge discovery and evidence synthesis: CoDiet has created the largest fully double annotated biomedical corpus in nutrition and non communicable diseases, coupled with advanced relation extraction pipelines based on open weight language models. This has resulted in a transparent, reproducible knowledge graph integrating more than 4,000 scientific publications, enabling more accurate and interpretable extraction of diet–disease mechanisms than existing literature based systems.

• Objective and scalable dietary assessment: The project advances dietary measurement beyond self report by validating complementary objective approaches, including urine NMR metabolomics, dried blood spot lipidomics and a passive dietary camera. In particular, scalable two dimensional image based portion estimation represents a step change compared with current digital dietary tools. The integration of biomarkers with sensor based data enables more reliable monitoring of real world dietary intake and short term metabolic responses.

• Multi omics integration and mechanistic biomarkers: CoDiet has generated one of Europe’s richest integrated datasets linking diet, microbiome composition, fatty acid profiles, metabolomics, genomics and cardiometabolic indicators. This unique resource enables the identification of mechanistic biomarkers that capture individual variability in metabolic response to diet, going beyond conventional epidemiological panels and supporting personalised prediction of NCD risk.

• Advanced machine learning methodology: The project has developed enhanced probabilistic and time series models tailored to high dimensional, small sample settings typical of deep phenotyping studies. Innovations include learning strategies informed by biological structure and knowledge graph based data augmentation. Notably, the integration of external scientific knowledge into Gaussian process priors for biomarker prediction opens new routes towards causally informed modelling in nutrition and metabolic health.

• Scalable computational infrastructure: CoDiet’s computational framework supports the execution of large scale multimodal workflows across distributed and federated environments, incorporating elastic resource allocation, GPU edge computing and optimised container scheduling. These capabilities, still uncommon in academic nutrition research, are essential for scaling AI based analysis of complex health data while respecting data governance constraints.

• Personalised dietary recommendations: The project has delivered an advanced personalised recommendation system that integrates behavioural science, multi modal data inputs and culturally adaptable communication strategies. By combining a participant facing mobile application, adaptive messaging, nutrition algorithms and user preference modelling, CoDiet moves beyond static dietary advice towards dynamic, data driven and user centred support for healthier eating.