PPML

PPML predicts phenotypic outcomes of phenylketonuria (PKU) from allelic genotypes using machine learning and graph-derived structural features.

Key Features:

• Machine Learning Framework: PPML employs a machine learning model to predict phenotypic outcomes from allelic genotypes in PKU.
• Mutation Analysis: The tool identifies and analyzes 235 different mutations across alleles and classifies them into classical PKU (cPKU), mild PKU (mPKU), and mild hyperphenylalaninemia (MHP).
• Graph-Based Features: Structural features of mutations and five node attributes derived from a PKU network graph are used as model inputs, with hub nodes c.728G>A (cPKU), c.721 (mPKU), and c.158G>A (MHP) emphasized for classification.
• Performance Metrics: Predictive performance is reported as area under the receiver operating characteristic curve (AUC): 0.832 for cPKU, 0.678 for mPKU, and 0.874 for MHP.

Scientific Applications:

• Phenotype prediction: Predicts PKU phenotypes from allelic genotypes to inform assessment of likely disease severity.
• Clinical interpretation: Supports interpretation of PAH gene variants’ likely phenotypic consequences to inform personalized treatment planning.

Methodology:

Models were trained on a dataset of 1,291 PKU patients and 623 distinct variants of the PAH gene, integrating structural mutation data with network graph properties to classify phenotypic outcomes.