Machine Learning-Enhanced Single-Particle Tracking for Rapid Screening of Tumor Immunomodulatory Drugs.

Abstract

The tumor microenvironment plays a critical role in tumor progression and immune response, with the extracellular matrix (ECM) regulating immune cell infiltration. However, the interplay between ECM dynamics and tumor immunity remains poorly understood, and current methods for evaluating immunomodulatory drugs are limited by scalability and cost. Here, we present a robust approach combining single-particle tracking (SPT) of quantum dots (QDs) with machine learning to characterize the extracellular space (ECS) in live tumor tissues and assess drug efficacy. Super-resolution images revealed dynamic changes in ECS geometry and ECM composition during tumor progression, validated by transcriptomics and flow cytometry. By extracting diffusion fingerprints from SPT data, we developed machine learning models to classify tumor stages and quantify immune infiltration with high accuracy. This platform enabled rapid ranking of immunomodulatory drugs based on their effects on ECS properties and immune responses, offering a cost-effective, and scalable method for advancing both drug discovery and personalized cancer immunotherapy.