Dual-site peroxidase-mimic graphdiyne-based colorimetric sensor arrays with machine learning for screening of multiple antibiotics

Abstract

The abuse of antibiotics poses a significant threat to both human health and the ecosystem, while the rapid screening of multiple antibiotics remains a challenge. We report the design of a dual-site peroxidase (POD)-mimic nanozyme comprising self-assembled hemin molecules and Cu²⁺ on graphdiyne (GDY) for screening of multiple antibiotics assisted with machine learning (ML). Cu ions can bond with π bonds and carbonyl groups on the surface of GDY, thereby enabling strong interface of hemin and GDY for an enhanced generation of hydroxyl radicals (^.OH). GDY/Hemin/Cu exhibits POD-like activity in wide pH conditions and temperatures ranging from 20 to 70 °C. With the assistance of ML, GDY/Hemin/Cu-based colorimetric sensor arrays demonstrate fast and accurate identification of multiple antibiotics including kanamycin, norfloxacin, ampicillin sodium, catechol and isoniazid. Theoretical calculation confirms that strong binding affinity enables specificity of the GDY/Hemin/Cu towards antibiotics. By employing support vector machine algorithm to assess antibiotic content, a high detection accuracy of 97.5% is achieved across 40 honey samples, underscoring the potential practical applications in screening of multiple antibiotics.