Machine Learning Model Integrating Nutrition, Genetics, and Microbiome for Recurrent Aphthous Ulcer Prediction

Abstract

Recurrent aphthous ulcers (RAU) are common inflammatory oral lesions characterized by repeated episodes of painful
ulceration, with susceptibility influenced by nutritional, genetic, microbial, immune, and environmental factors. Conventional
approaches to RAU prediction primarily depend on clinical history and identifiable risk factors, limiting their ability to
capture complex interactions among biological determinants. This paper proposes a machine learning model that integrates
nutritional, genetic, and microbiome data to improve individualized prediction of RAU recurrence. Nutritional variables may
include dietary patterns, micronutrient intake, and nutritional deficiencies, while genetic features may encompass immunerelated
susceptibility markers and inflammatory response variants. Oral and gut microbiome characteristics, including
microbial diversity, abundance, and dysbiosis-associated patterns, can provide additional biological predictors. Machine
learning algorithms can combine these heterogeneous datasets to identify nonlinear relationships and generate individualized
risk scores. Explainable artificial intelligence may further clarify the relative contribution of dietary, genetic, and microbial
factors to predicted risk. Such a multimodal framework could support early identification of susceptible individuals and
enable personalized preventive strategies, including nutritional modification and targeted oral-health interventions. Future
validation using large longitudinal cohorts, standardized biological measurements, and diverse populations will be essential
for establishing clinical reliability and practical implementation.