Precision nutrition combines episodic and error-prone inputs, such as dietary records, with biomarkers, multi-omics, microbiome, wearable sensors and behavioural information. Data volume alone does not ensure reliable recommendations.

The Nature Communications perspective highlights harmonisation, missing data, database differences, technical variation, cohort representativeness, interpretability and overfitting.

The authors identify external and prospective validation as the reference: testing whether an AI-guided intervention actually improves biomarkers or behaviour while maintaining performance across populations, sites and time periods.

For industry, this changes the question. It is not enough to demonstrate that a platform personalises; it must disclose the data used, how they were measured, the population for which the model is valid, the outcome affected and the added value over simpler interventions.

Why it matters

For functional ingredients and digital services, advantage is shifting from the promise of personalisation to documented data quality, performance, usefulness and limitations.

Evidence limit

A peer-reviewed methodological perspective, not an efficacy trial or proof that currently available platforms automatically improve biomarkers or behaviour.

Editorial note. This content is intended for industry professionals. It does not constitute medical advice, therapeutic guidance or regulatory advice. Read our editorial method.

Read the original source · Nature Communications ↗