Machine Learning Assisted Nanofluidic Array for Multiprotein Detection

Abstract

Solid-state nanopore and nanochannel biosensors have revolutionized protein detection by offering label-free, highly sensitive analyses. Traditional sensing systems (1st and 2nd stages) primarily focus on inner wall (IW) interactions, facing challenges such as complex preparation processes, variable protein entry angles, and conformational changes, leading to irregular detection events. To address these limitations, recent advancements (3rd stage) have shifted toward outer surface (OS) functionalization but are constrained by single-protein recognition models. Herein, we show a machine learning assisted nanofluidic array (MANY) sensing system (4th stage) that integrates a supervised dimensionality reduction strategy with photoresponsive MoS₂ nanofluidic array functionalized with nonspecific functional elements (FEarray) at the OS. This approach serves as a proof-of-concept for label-free, probe-free detection of multiple proteins with 100% accuracy, highlighting its significant potential for rapid diagnostics in future disease detection applications.