Building an AI-Powered Educational Tool for Exploring Microbial Relationships in Parkinson’s Disease

This paper presents the Neurobiome Navigator, an AI-powered, highly interactive, and easily navigable application designed to help users explore the complex relationships between the human microbiome and Parkinson’s disease (PD). The app focuses on the gut microbiome and the oral microbiome, known to have a strong relationship with PD, as well as impulse control disorders (ICD), a significant non-motor symptom of PD. The system integrates the MINERVA (Microbiome Network Research and Visualization Atlas) knowledge graph with supplemental scientific literature to deliver evidence-based insights. Built using an AI agentic workflow including Streamlit, Neo4j, and LangChain, the application enables users to submit structured survey responses (e.g., oral health, impulse control) or free-text queries. Data retrieval methods include Neo4j graph queries and semantic vector search. Retrieved content is synthesized by a large language model (GPT-4o) and an agentic pipeline to generate personalized, non-clinical suggestions. By bridging advanced AI capabilities with an accessible interface, the Neurobiome Navigator aims to empower users to make informed lifestyle decisions that can potentially improve quality of life, particularly in managing non-motor symptoms of PD. The platform enhances understanding of microbiome-PD connections by presenting complex scientific relationships in an interactive, visually engaging, and easily navigable format. Through personalized insights, dynamic charts, and exploratory tools, it transforms dense biomedical data into actionable, user-friendly guidance, making the learning process both informative and enjoyable.