The Power of Artificial Intelligence for Phenylketonuria

Principal Investigator:
Malihe Aliasgari

Co-PIs:
Yousef Nejatbakhsh, Roya Nejatbakhsh, Haoru Li, Ying Zhang

Abstract:
In this work we present the first and novel health-aware diet system for individuals with phenylalanine hydroxylase (PAH) deficiency, that aligns with their preferences and restricts the protein based on the level of phenylalanine in their blood test.

Description of Research:
PKU is a genetic metabolic disorder. Patients with the conditions are unable to convert the amino acid phenylalanine into tyrosine. Deficiency of this enzyme leads to an increased production of phenylketone bodies (hence phenylketonuria) resulting in high phenylalanine levels in the blood and brain. Without treatment, the conditions can lead to brain damage, intellectual disabilities, seizures, and behavioral or psychiatric disorders. The main treatment for PKUs includes a strict lifetime diet with a very limited intake of protein for the rest of their lives to make sure to get enough essential protein and nutrients that are essential for growth and general health. The diet should be started as soon after birth as possible. However, the treatment regime is complex, costly, and often difficult to maintain for the long term. A food recommender system could be a valuable tool for PKU patients, helping them to quickly and easily identify foods that are safe for them to eat. The system would use machine learning to learn about the Phenyl content of different foods and the individual dietary needs of each patient. It would then generate personalized recommendations for foods that are low in Phenyl and meet the patient’s other dietary needs, such as calorie requirements, and food protein. This project proposes to develop and evaluate a healthy and time-aware food recommender system for PKU patients from birth to adulthood, using a combination of machine learning techniques, including content-based filtering, a hybrid system, a rule-based system, and a large language model. The system will be evaluated in a clinical trial with PKU patients to assess its accuracy, usability, and impact on dietary adherence and health outcomes.