Machine learning-assisted biosensor for microRNA analysis based on RCA-mediated hemin/G-quadruplex and thiol-functionalized red carbon dots.

Tumor biomarkers play a critical role in the early detection of cancer. Therefore, there is an urgent need for rapid, user-friendly, and precise miRNA detection methods suitable for resource-limited settings. In this study, a label-free fluorescence biosensor was developed, combining rolling circle amplification (RCA) with red-emitting carbon dots (R-CDs) for the ultrasensitive and specific identification of miRNA let-7a. Initially, R-CDs with long-wavelength emission were synthesized and functionalized with thiol groups via an amidation reaction. A template strand was designed containing sequences complementary to both the G-quadruplex and miRNA let-7a, along with a specific Nt.BbvCI nicking enzyme recognition site. Using miRNA let-7a as a primer, the RCA process was initiated. The Nt.BbvCI enzyme triggered the generation of numerous DNA fragments, which subsequently acted as primers to initiate new RCA cycles. The resulting G-quadruplexes bound to hemin, forming DNAzymes with peroxidase-like activity that decomposed HO into hydroxyl radicals (・OH). These radicals oxidized the thiol groups on the R-CDs-SH, forming disulfide bonds (-S-S-) that led to oxidative aggregation and fluorescence quenching. This strategy achieved an ultralow detection limit of 0.16 fM. This biosensor, which integrates R-CDs, a label-free approach, and nicking-enhanced RCA, holds great promise for early cancer diagnostics.