Developing an augmented nutrient profiling system in the perspective of healthy and sustainable diets

Nutrient profiling systems (NPSs) are widely used to guide public health policies. However, the urgent threat posed to human life by climate change highlights the need to expand the concept of healthy diets to include sustainability. Thus, this study aimed at developing an NPS that integrates nutritional quality, sustainability and food processing by creating a machine learning (ML)-based artificial neural network. ML-based NPSs showed strong Spearman's correlations with established healthy and sustainable diets (Mediterranean diet: rho = 0.891, <i>p</i> < 0.001; Eat-Lancet reference diet: rho = 0.783, <i>p</i> < 0.001), as well as a composite environmental sustainability index (rho = 0.710, <i>p</i> < 0.001). In contrast, ML-based NPSs showed weak correlations with NOVA classification, revealing an inverse relationship between the Sustainability Index and NOVA. These findings suggest that deep learning methods can effectively balance multiple dimensions in NPS design. In conclusion, the proposed method can classify any food product and offer guidance on appropriate consumption frequencies to consumers, promoting healthier and more sustainable diets.