Machine learning-assisted renewable and polarity-switchable photoelectrochemical biosensor for circRNA intelligent diagnosis.

Herein, a renewable polarity-switchable PEC biosensor is reported for a highly sensitive and selective assay of circRNA in human whole blood, and machine learning is exploited to assist in the circRNA intelligent diagnosis. The target circRNA-DNA3 binary structure can hybridize competitively with DNA2 in triple helix molecular to generate three-way junction probe, and Exo Ⅲ specifically cleaves DNA2 to release the circRNA-DNA3 binary complex for signal cyclic amplification. Then, Cu2O nanospheres are introduced into PEC platform, leading to the switching from anodic to cathodic photocurrents. Interestingly, biotin can competitively bind to Cu2O-SA, making the ITO/CdS/T-COF/CS/DNA1 electrode reusable for circRNA analysis. The built PEC biosensor exhibits a low detection limit (7.6 aM), excellent selectivity and satisfactory renewability. Moreover, the developed PEC biosensor for human whole blood circSATB2 assay can effectively distinguish lung cancer patients from healthy individuals (P < 0.001). Importantly, the machine learning is adopted to explore the potential pattern hidden in PEC data, and the accuracy, sensitivity and specificity of circRNA intelligent diagnosis all reach 100 %. Machine learning-assisted renewable polarity-switchable PEC biosensor provides a new approach for circRNA analysis and early intelligent diagnosis of cancer.