Machine Learning Assisted Biosensing for Early Inflammation Modulation

Abstract

Full comprehension of how the immune system works is crucial for precise prognosis and effective treatment of immune-related diseases. The practice of precision immune profiling and monitoring is crucial for personalized monitoring, disease evolving tracking, and drug efficacy screening, offering prompt care for urgent immune response. However, challenges remain in precision immune profiling, including limited sample availability, sensitivity, high throughput screening and spatiotemporal resolution. This dissertation focuses on the engineering and development of nanoplasmonic immunoassays for acute immune disease monitoring. Specifically, rational designed peptide aptamers as probes, are introduced into immunoassay to ensure the sensitive and specific signal for immune profiling. Additionally, multiple nanoplasmonic nanoparticles enable high-throughput, multiplexed, in-situ detection while creating a sepsis organoid microenvironment. Simultaneously, a machine-learning based algorithm was developed to quickly analyze the signals and interpret the data to reduce the assay time. In the first project, we developed an ultrasensitive nanoplasmonic digital immunoassay that integrates rationally engineered antibody-derived peptide aptamers (ADPAs), plasmonic gold nanospheres for digital dark-field imaging, and convolutional neural network signal quantification. This platform enables ultrasensitive cytokine detection down to tens of fg/mL, with quantitative coverage across the tested concentration range spanning approximately six orders of magnitude using only microliter-scale sample volumes. We applied this technology to a physiologically relevant in vitro model of CAR T-cell induced cytokine release, enabling high-frequency cytokine monitoring at early time points previously inaccessible to conventional assays. In the second project, we developed a micropillar immunoassay, integrated with multiple types of nanoparticles, facilitates digital, real-time, in situ, multiplexed immune profiling within a microenvironment on-chip. The versatile microenvironment on-chip designs show promise for applications in complex cancer-immune microenvironment analyses, tumor organoid studies, and high-throughput single-cell analyses.