Machine learning provides a step toward a stable electrochemical biosensor for histamine measurement in meat samples

Conventional electrochemical histamine biosensors face limitations due to short enzyme stability and restricted linear detection ranges. This study introduces a biosensing device that integrates electrochemistry with machine learning to address these challenges and improve sensitivity, stability, and long-term performance. The electrochemical unit was designed using diamine oxidase co-immobilized with potassium hexacyanoferrate as an electron mediator, combined with a chitosan-gold nanoparticle cryogel to enhance conductivity, all applied to a screen-printed carbon electrode. This configuration enabled a wide linear detection range of 0.5–90 ppm and a low detection limit of 0.1 ppm via chronoamperometry at +200 mV vs. silver-ink pseudo-reference electrode. For the machine learning component, various models were trained using chronoamperometric current data from samples containing various histamine concentrations (0–200 ppm), electrode storage temperatures (4–8 °C), and electrode lifespans (0–45 days). The artificial neural network optimized with genetic algorithms accurately predicted histamine concentrations even in nonlinear ranges, achieving strong performance (R 2 = 0.95, RMSE = 13.2 ppm), while maintaining reliability for up to 40 days for a storage temperature within 4–8 °C. When validated against ELISA in beef and trout samples, the biosensor showed superior precision. The combined electrochemical–machine learning approach offers a rapid and stable tool for meat spoilage detection. • A histamine biosensor with a linear range of 0.5-90 ppm and LoD of 0.1 ppm • A model based on machine learning to improve the biosensor stability up to 40 days • A working range of 0-200 ppm using the machine learning decision-making unit • Better precision compared with ELISA in measuring histamine in beef and trout