Allelic phenotype prediction of phenylketonuria based on the machine learning method

Area of Science:

• Genetics
• Bioinformatics
• Machine Learning

Background:

• Phenylketonuria (PKU) is a genetic disorder caused by mutations in the phenylalanine hydroxylase (PAH) gene.
• Predicting the specific phenotype of PKU based on allelic genotype is crucial for effective management.

Purpose of the Study:

• To develop and validate a machine learning framework for predicting PKU phenotypes from allelic genotypes.
• To classify PKU phenotypes into classical PKU (cPKU), mild PKU (mPKU), and mild hyperphenylalaninemia (MHP).

Main Methods:

• Utilized a dataset of 1291 PKU patients with 623 variants for training.
• Developed a machine learning framework (PPML) incorporating mutation structure and PKU network graph properties.
• Extracted features from mutation structure and network graph attributes, including three hub nodes for classification.

Main Results:

• Identified 235 mutations and 623 allelic genotypes.
• The PPML model achieved high predictive accuracy: AUC=0.832 for cPKU, AUC=0.678 for mPKU, and AUC=0.874 for MHP.
• Demonstrated PPML as a powerful tool for predicting PKU allelic phenotypes.

Conclusions:

• PPML effectively predicts PKU allelic phenotypes, aiding in genetic counseling for PKU families.
• An online web version of PPML is available for PKU phenotype prediction: http://www.bioinfogenetics.info/PPML/.