RDAD

RDAD: Phenotype-Based Rare Disease Diagnostic Modeling System

RDAD integrates clinical phenotypic data with phenotypic similarity assessment and machine learning algorithms to prioritize rare Mendelian disease candidates and generate ranked diagnostic predictions.

Key Features:

• Dual-Method Framework: Combines phenotypic similarity evaluation with machine learning to construct four diagnostic models.
• Disease Prioritization: Generates ranked predictions listing the top 10 candidate rare diseases per clinical case.
• Model Performance: Achieves ≥98% precision and up to 95% recall based on validation with RAMEDIS clinical records.
• Pattern Recognition: Machine learning models detect complex phenotypic patterns associated with specific rare diseases.

Scientific Applications:

• Rare Mendelian Disease Diagnosis: Supports identification of molecularly unresolved rare diseases using phenotype-driven computational prioritization.
• Clinical Decision Support: Provides evidence-based disease candidate rankings to guide diagnostic evaluation and personalized medicine.

Methodology:

RDAD constructs four diagnostic models using clinical phenotype data. It applies phenotypic similarity scoring and machine learning algorithms to identify candidate rare Mendelian diseases and outputs the top 10 ranked predictions for each evaluated case. Model performance was validated using real-world clinical records from RAMEDIS.