Project Summary/Abstract Neurofibromatosis (NF) encompasses a set of complex genetic disorders that affect almost every organ system and increase risk for the development of benign and malignant central and peripheral nervous system tumors. Of the three types of NF, Neurofibromatosis Type 1 (NF1) is the most prevalent occurring in approximately 1 in every 3,000 births without predilection for race, sex, or ethnicity. While NF1 is inherited in a fully penetrant autosomal dominant manner, there is wide inter-individual variability with respect to clinical features and their impact on patient morbidity. Clinical heterogeneity is a pervasive challenge for clinicians and families, as the management of children and adults with NF1 remains largely reactive, without reliable biomarkers or predictive models for early risk stratification and/or prognostic assessment at the time of diagnosis. Traditional approaches, which focus on identifying a single clinical or biological marker that can be measured and used to assess disease risk or trajectory in NF1, have achieved limited success and have hindered progress in the development of precision medicine for NF1-affected individuals. In response to these challenges, and with the opportunity to improve the care of individuals with NF1, we aim to verify and validate an alternative and generalizable approach for developing artificial intelligence (AI)-based clinical decision support tools for NF1 sub-phenotypes, implemented and evaluated in a comparative manner across two clinical sites. Our proposed project will first generate a multi-scale data set using a text-mining based clinical phenotyping algorithm to integrate and harmonize data from multiple sources such as clinical databases, structured electronic health records, and unstructured clinical notes. Secondly, we will develop AI-based pipelines capable of generating predictive models and tools to identify disease risk for three critical NF1 sub- phenotypes (OPGs, scoliosis, and ADHD). We will then evaluate the models for quantitative accuracy and clinical actionability at the point of care with the help of NF1 clinicians. Finally, we will validate these methods and models across multiple sites, so that we can better understand the challenges to generalizing and transporting such predictive models based across different healthcare systems, environments, and populations. We anticipate that the use of artificial intelligence techniques in order to study NF1-specific sub-phenotypes at two different sites will yield novel and potentially clinically-actionable and generalizable insights concerning the precision diagnosis and care of individuals with NF1, with broader applicability across a spectrum of similarly complex disease-states.