"Three-in-one" Analysis of Proteinuria for Disease Diagnosis through Multifunctional Nanoparticles and Machine Learning.

Abstract

Urinalysis is one of the predominant tools for clinical testing owing to the abundant composition, sufficient volume, and non-invasive acquisition of urine. As a critical component of routine urinalysis, urine protein testing measures the levels and types of proteins, enabling the early diagnosis of diseases. Traditional methods require three separate steps including strip testing, protein/creatinine ratio measurement, and electrophoresis respectively to achieve qualitative, quantitative, and classification analyses of proteins in urine with long time and cumbersome operations. Herein, this work demonstrates a "three-in-one" protocol to analyze the urine composition by combining multifunctional nanoparticles with machine learning. This work constructs a sensor array to analyze proteinuria by employing nanoparticles with unique optical properties, outstanding catalytic activity, diverse composition, and tunable structure as probes. Different proteins interacted with nanoprobes differently and are classified by this sensor array based on their physicochemical heterogeneities. With the aid of machine learning, the urine composition is precisely detected for the diagnosis of bladder cancer. This protocol enables quantification and classification of 5 proteinuria in 10 min without any tedious pretreatment, showing proimise for the comprehensive analysis of body fluid for early disease diagnosis.