A Machine Learning-Driven Electrophysiological Platform for Real-Time Tumor-Neural Interaction Analysis and Modulation

Abstract

Abstract Neural-tumor electrophysiology—marked by pathological membrane potentials and ion channel dysregulation—emerges as actionable targets to curb tumor aggression. Yet, how neural-driven bioelectrical crosstalk dynamically regulates tumors within functional circuits remains elusive, demanding tools for real-time interaction decoding. Here, we present a machine learning-driven electrophysiological platform that integrates custom microfluidics with real-time decoding of complex neural-tumor signal dynamics. Our findings show that glioma cells selectively hijack specific subsets of neural signals, reshaping waveform properties and synchronizing their firing events with neural activity. This dynamic interaction plays a critical role in boosting glioma invasiveness, as tumor cells harness neural activity to promote their progression. Notably, targeted stimulation of glioma cells with these hijacked signal patterns—without direct neural involvement—is sufficient to induce hyper-invasive behavior, emphasizing the role of these electrical cues as drivers of tumor aggression.