Biosensor fotoelectroquímico asistido por aprendizaje automático basado en nanocables de ADN-AgNC para el diagnóstico...

Exosomal long non-coding RNA (lncRNA) as a burgeoning biomarker, its abnormal expression is closely related to the progression of cancer. Therefore, the development of sensitive exosomal lncRNA detection methods has significant effect in promoting the early diagnosis of cancers. Herein, a novel signal-on photoelectrochemical (PEC) biosensor is reported for the highly sensitive detection of exosomal lncRNA combined porous COF/TiO₂ nanospheres and terminal deoxynucleotidyl transferase (TdT)-triggered poly(C) with silver nanoclusters (DNA-AgCNs) nanowires, using machine learning to assist in cancer intelligent diagnosis.

The as-prepared COF/TiO₂ nanospheres as photoelectrode materials with high specific surface area and porosity can carry abundant tyrosine (Tyr). Then, target lncRNA HOTAIR can bind with phosphate groups at the 3'-end of DNA (pDNA) to expose the porous sites to adsorb Tyr. Subsequently, TdT can catalyze deoxycytidine triphosphate (dCTP) cyclic amplification to in-situ generate abundant DNA-AgCNs nanowires, which can act as a bridge for electron transfer, enhancing the photocurrent of COF/TiO₂ nanospheres. The developed PEC biosensor achieves a wide range from 500 aM to 10 pM with a low detection limit of 116 aM. Importantly, this machine learning is employed to probe the hidden potential pattern in the developed PEC data, and machine learning for cancer intelligent diagnosis can achieve 85.7 % accuracy, 100 % sensitivity and 80.0 % specificity.

Machine learning-assisted PEC biosensor based on DNA-AgNCs nanowires and porous COF/TiO₂ nanospheres can effectively distinguish the expressions of lncRNA HOTAIR in plasma exosomes from healthy people and cancer patients, which not only significantly enhances the sensitivity and accuracy of cancer diagnosis, but also provides a great application prospect in the early diagnosis of lncRNA-related cancer.