Size-Coded Hydrogel Microbeads for Extraction-Free Serum Multi-miRNAs Quantifications with Machine-Learning-Aided Lung Cancer Subtypes Classification

Classifying lung cancer subtypes, which are characterized by multi-microRNAs (miRNAs) upregulation, is important for therapy and prognosis evaluation. Liquid biopsy is a promising approach, but the pretreatment of RNA extraction is labor-intensive and impairs accuracy. Here we develop size-coded hydrogel microbeads for extraction-free quantification of miR-21, miR-205, and miR-375 directly from serum. The hydrogel microbead is immobilized with an miRNA capture probe, which well retains target miRNA and provides good nonfouling capability for nonspecific biomolecules in serum. The porous structure of microbeads allows efficient DNA cascade amplification reaction and generates a fluorescence signal. The microbeads are clustered into three groups according to size via flow cytometry sorting, and the group fluorescence is integrated for the corresponding miRNA quantification. With machine-learning-assisted data analysis, it achieves good lung cancer diagnosis accuracy and 80% accuracy for subtype classification for 108 serum samples, including lung cancer patients and healthy controls.