A target amplification-free CRISPR biosensor with digital microfluidic processing for microRNA profiling in AI-enhanced pediatric sepsis diagnosis.

Abstract

MicroRNA signatures provide an informative readout of host-response dysregulation in pediatric sepsis, but their clinical translation remains limited by low target abundance and multiplexing requirements. Here, a target amplification-free CRISPR biosensing platform was established. The sensing mechanism couples Cas13a target recognition to promoter release, template reconstruction, and transcription-driven Cas13a reactivation, enabling fluorescence amplification through internal molecular cycling without amplifying the target miRNAs. Then, an electrowetting-based digital microfluidic chip was designed to compartmentalize parallel reactions and automate droplet generation, reaction assembly, isothermal incubation, magnetic separation, and fluorescence readout. Under optimized conditions, four sepsis-associated biomarkers (miR-15a, miR-21, miR-146a, and miR-155) and the reference control (snRNA U44) were detected at attomolar levels (expected mean input of 1.67 molecules per reaction) with good specificity and minimal inter-droplet interference. The resulting miRNA profiles were then used as input features for machine learning-enhanced sepsis classification and risk assessment. In 252 clinical samples, the resulting model achieved a test-set AUC of 0.973 and an overall accuracy of 92.1%. The model-derived risk score was associated with the Phoenix Sepsis Score and, in a representative longitudinal case, showed preliminary potential to reflect disease evolution. Overall, this platform enables automated analysis of disease-associated miRNA signatures and demonstrates its potential for in-vitro diagnosis of pediatric sepsis.