Abstract
Personalized nutrition is becoming increasingly important as we discover just how differently each person's body processes nutrients based on their genetics, physical condition, and even mental well-being. Traditional dietary guidelines take a one-size fits-all approach and miss the complicated ways these factors interact with each other. Our study tackles this problem by creating a multi-modal deep learning framework that can combine different types of biological and behavioral data to predict personalized nutrient needs. Our model takes in four main types of data: basic health information (like age, BMI, blood pressure, cholesterol levels, and whether someone has diabetes), mental health factors (stress levels, sleep quality, and mood), genomic SNP sequences, and protein-folding patterns that come from genetic variants. We designed separate neural network branches to handle each type of data - dense layers work through the health metrics, a bidirectional LSTM processes the mental health data over time, a Transformer encoder analyzes the genomic information, and a stacked bidirectional LSTM tackles the protein-folding sequences. All these different representations get combined and fed through additional dense layers to ultimately predict eight nutritional targets: total grams, energy, protein, carbohydrates, sugars, fiber, fat, and cholesterol. The system was trained on a large integrated dataset of 73,778 dietary records from 4,983 individuals, combining NHANES 2021–2023 with SNP and protein-fold data from the 1000 Genomes Project. The multi-modal design enables the model to learn complex connections across biological and behavioral domains, offering richer insights into nutrient patterns than traditional single-source methods. This work demonstrates that fusing genomic, proteomic, physiological, and psychological information can substantially enhance personalized nutrition analysis. This approach gives us a solid starting point for building smarter dietary recommendation systems that account for each person's specific genetic makeup and health situation.