New Publication in PLOS Digital Health: A Few Smart Swaps for Healthier, Cheaper Meals

Trevor Chan and Prof. Ilias Tagkopoulos published Translating dietary standards into healthy meals with few-ingredient substitutions in PLOS Digital Health on 5/28/2026.

About this work: What if eating healthier meant changing only one or two things on your plate? This work introduces an end-to-end framework that turns USDA dietary standards into complete, realistic meals while asking for as few ingredient changes as possible. Drawing on 135,491 meals from national intake data, the authors distilled eating patterns into 34 interpretable meal archetypes and trained a generative model, paired with a portion predictor, to hit per-meal nutritional targets. The generated meals cut absolute deviation from recommended intake by a median of 47.0% while still looking like food people actually eat, and just one to three substitutions yielded meals that were on average 10% more nutritious at 19 to 32% lower cost. The approach also outperformed GPT-4o on dietary-guideline compliance, pointing toward practical, low-friction tools for healthier eating.

Reference: Chan, Trevor, and Ilias Tagkopoulos. “Translating dietary standards into healthy meals with few-ingredient substitutions.” PLOS Digital Health (2026). doi: 10.1371/journal.pdig.0001367