PNA/SWCNT hybrid electrochemical platform optimized by machine learning for species-specific quantification of Bacillus cereus.

The rapid detection of foodborne pathogens is paramount for public health. Herein, we report a first machine-learning (ML)-optimized peptide nucleic acid (PNA)-based electrochemical biosensor for sensitive and specific detection of Bacillus cereus. The biosensor platform was constructed by functionalizing a screen-printed carbon electrode with carboxylated single-wall carbon nanotubes (SWCNT-COOH) to enhance conductivity and probe loading. A specific PNA probe, complementary to the motB gene, was then immobilized to capture the target DNA. A key innovation was the use of an ML model to systematically optimize critical fabrication and operational parameters, including SWCNT concentration, PNA density, and applied potential, thereby drastically streamlining the development process. Under optimized conditions, the biosensor detected target DNA via a concentration-dependent decrease in the ferri/ferrocyanide amperometric signal (i-t) with a detection time of 180 s. It achieved a low detection limit (LOD) of 0.3 ng/μL, exhibited excellent stability for two weeks, and demonstrated reliable performance in complex matrices, with recovery of 78-105% in drinking water and 90-111% in milk. This work establishes a robust, ML-guided biosensing strategy that combines PNA's affinity with the enhanced transduction of nanomaterials, providing a powerful and practical platform for food safety monitoring.