Random forest classifier improving phenylketonuria screening performance in two Chinese populations

Area of Science:

• Biochemistry
• Genetics
• Medical Diagnostics

Background:

• Phenylketonuria (PKU) is a genetic metabolic disorder affecting amino acid metabolism.
• Untreated PKU can cause severe developmental harm in newborns and children.
• Early diagnosis and intervention are crucial for preventing disease progression.

Purpose of the Study:

• To develop and validate a novel screening model for Phenylketonuria (PKU).
• To enhance PKU screening performance using machine learning.
• To compare the efficacy of a random forest classifier (RFC) against other models.

Main Methods:

• Development of a PKU screening model utilizing a random forest classifier (RFC).
• Validation of the RFC model on a diverse dataset, including two Chinese populations.
• Comparative analysis of RFC against traditional logistic regression and other machine learning models.

Main Results:

• The RFC model demonstrated excellent sensitivity, false positive rate (FPR), and positive predictive value (PPV).
• RFC outperformed other classification models, including logistic regression, in PKU screening.
• Consistent high performance was observed across validation and testing datasets.

Conclusions:

• The random forest classifier (RFC) offers a promising advancement for neonatal PKU screening.
• RFC provides a robust and accurate method for early detection of Phenylketonuria.
• This model has the potential to significantly improve PKU diagnostic capabilities.