Machine Learning Identifies Diet Patterns Associated with Lower Mortality

A new study published in npj Science of Food on August 7, 2026, has developed a machine learning-based diet score that identifies eating patterns associated with reduced aging-related mortality. The MYTH Diet study analysed 191,689 UK Biobank participants with a mean age of 56.1 years, of whom 55.1% were female.

The MYTH Diet Score: Design and Components

Researchers grouped 206 food items into 34 food groups and identified 18 that were significantly associated with mortality after false discovery rate correction. A LightGBM model was used to rank food group importance, and the final MYTH Diet score comprises ten components on a scale of 0 to 10.

The scoring system encourages poultry, nuts, and refined grains, and discourages processed meat, red meat, butter, and sweetened beverages. Legumes, coffee, and alcohol are scored within a moderate intake range.

Real-World Outcomes: UK Biobank Data

Over a median follow-up of 12.2 years in the UK Biobank cohort, 13,652 participants (7.1%) experienced aging-related death. Higher MYTH Diet scores were associated with lower aging-related mortality in the UK Biobank, with a hazard ratio of 0.79 when comparing the top versus bottom quartile.

External Validation and Comparative Performance

When tested in the external NHANES validation cohort, higher MYTH Diet scores were associated with lower aging-related mortality with a hazard ratio of 0.68. However, the MYTH Diet performed comparably to, rather than consistently better than, the Healthy Eating Index, which showed a stronger association when modelled continuously in NHANES.

Biological Mechanisms: Protein Signatures and Aging Markers

A 50-protein signature of the MYTH Diet score explained an estimated 26.7% of its association with mortality, with TNFRSF4 and CD74 among the leading protein mediators. Higher MYTH Diet scores were also associated with slower aging on plasma protein-derived clocks for the lung, pancreas, kidney, liver, and arteries.

Broader Disease Associations

Higher MYTH Diet scores were associated with lower risk of 15 of 49 age-related diseases tested, suggesting broader protective effects beyond mortality alone.

Limitations and Generalisability

The MYTH Diet model’s discriminative performance was modest, with an AUC of 0.689. Both the UK Biobank and NHANES cohorts are predominantly of Western or White European background, limiting generalisability to diverse populations.

Source: 2 Minute Medicine