Improving the Diagnosis of Phenylketonuria by Using a Machine Learning-Based Screening Model of Neonatal MRM Data.

Zhixing Zhu,Xiaoshu Cai,Jing Guo,Yangmin Wang,Guoli Tian,Jianlei Gu,Hui Lu,Georgi Z Genchev

doi:10.3389/fmolb.2020.00115

Zhixing Zhu, Xiaoshu Cai + Show 6 more

Open Access

https://doi.org/10.3389/fmolb.2020.00115

Copy DOI

Abstract

Phenylketonuria (PKU) is a common genetic metabolic disorder that affects the infant's nerve development and manifests as abnormal behavior and developmental delay as the child grows. Currently, a triple–quadrupole mass spectrometer (TQ-MS) is a common high-accuracy clinical PKU screening method. However, there is high false-positive rate associated with this modality, and its reduction can provide a diagnostic and economic benefit to both pediatric patients and health providers. Machine learning methods have the advantage of utilizing high-dimensional and complex features, which can be obtained from the patient's metabolic patterns and interrogated for clinically relevant knowledge. In this study, using TQ-MS screening data of more than 600,000 patients collected at the Newborn Screening Center of Shanghai Children's Hospital, we derived a dataset containing 256 PKU-suspected cases. We then developed a machine learning logistic regression analysis model with the aim to minimize false-positive rates in the results of the initial PKU test. The model attained a 95–100% sensitivity, the specificity was improved 53.14%, and positive predictive value increased from 19.14 to 32.16%. Our study shows that machine learning models may be used as a pediatric diagnosis aid tool to reduce the number of suspected cases and to help eliminate patient recall. Our study can serve as a future reference for the selection and evaluation of computational screening methods.

Highlights

Phenylketonuria (OMIM:261600) (PKU) is a common inborn genetic metabolic disorder, which affects the infant’s neural development and manifests as abnormal behavior and developmental delay, which becomes apparent as the child grows
We applied multiple LRA models, a supervised machine learning algorithm for constructing method applicable in pediatric diagnostic screening in PKU utilizing highdimensional metabolic data. These models achieved performance of guaranteed Sn >95%, achieved AUC higher than 90%, and improved Sp and positive predictive value (PPV) on both the test set and independent test set
We reported a new marker of PKU—Met/Phe

Summary

Introduction

Phenylketonuria (OMIM:261600) (PKU) is a common inborn genetic metabolic disorder, which affects the infant’s neural development and manifests as abnormal behavior and developmental delay, which becomes apparent as the child grows. In most countries around the world, PKU diagnosis is performed by evaluating phenylalanine (Phe) and tyrosine (Tyr) levels in neonatal dry blood spots (DBSs) by MS/MS (Blau et al, 2014). A triple–quadrupole mass spectrometer [multiple reaction monitoring (MRM)] is used to measure 44 metabolites in neonatal blood samples, and in clinical practice, more than 30 types of genetic metabolic diseases (including PKU) are diagnosed by using these biomarkers. An estimated 13% positive predictive value (PPV) in PKU neonatal screening (Zhang et al, 2019) indicates that the pathophysiology of the disease is not uniquely driven by elevated level of Phe and Tyr. the large number of metabolites captured in experimental MRM data creates inherent complexity, which makes the overall meaningful signals non-trivial to assess by a manual process for a sizable patient population

Objectives

Methods

Results

Discussion

Conclusion