Predictive Modeling for Clinical Features Associated With Neurofibromatosis Type 1

Area of Science:

• Genetics and Genomics
• Pediatric Neurology
• Computational Biology

Background:

• Neurofibromatosis type 1 (NF1) is a complex genetic disorder with variable clinical manifestations.
• Optic pathway glioma (OPG) and attention-deficit/hyperactivity disorder (ADHD) are common complications in pediatric NF1.
• Predictive models for these complications are needed for early intervention and improved management.

Purpose of the Study:

• To conduct a longitudinal analysis of clinical features in NF1 patients.
• To explore the feasibility of using machine learning to predict OPG and ADHD in a pediatric NF1 cohort.
• To identify demographic and clinical predictors for OPG and ADHD in NF1.

Main Methods:

• Retrospective analysis of a curated NF1 clinical registry and electronic health records (EHR).
• Development of machine learning models, including gradient boosting classification.
• Analysis of data from 578 pediatric NF1 patients.

Main Results:

• White children had a higher likelihood of developing OPG (OR: 2.11).
• Males were more likely to be diagnosed with ADHD (OR: 1.90) and at an earlier age.
• Machine learning models achieved an AUROC of 0.74 for ADHD and 0.82 for OPG prediction.

Conclusions:

• Clinical and EHR data can effectively identify patterns in NF1 semiology.
• Machine learning models show promise for developing predictive tools for NF1-related complications.
• These predictive models can aid in risk stratification and disease management for NF1 patients.