Machine learning integrated MIL-88B@ZIF-67 built-in electric field heterojunction nanozyme for enhanced colorimetric detection of ascorbic acid and alkaline phosphatase.

A MIL-88 B@ZIF-67 heterojunction nanozyme was rationally designed by adopting a MOF-on-MOF assembly strategy to boost catalytic performance through interfacial electronic structure modulation. Owing to the Fermi level difference between ZIF-67 and MIL-88 B, spontaneous electron transfer from ZIF-67 to MIL-88 B occurs upon interfacial contact, leading to the establishment of a built-in electric field that promotes charge separation and accelerates electron migration. As a result, MIL-88 B@ZIF-67 exhibits significantly enhanced peroxidase-like activity toward the oxidation of 3,3',5,5'-tetramethylbenzidine (TMB) in the presence of H2O2. Based on this feature, the developed colorimetric sensing platform enabled sensitive detection of ascorbic acid (AA) within a limit of detection (LOD) of 0.65 μM and a linearity of 1-82 μM. Furthermore, a cascade sensing system for alkaline phosphatase (ALP) was developed by exploiting its hydrolysis of ascorbic acid phosphate (AAP) into AA, affording a linear range from 0.5 U·L-1 to 68 U L-1 and a LOD of 0.23 U L-1. Machine learning analysis shows that the Random Forest-Support Vector Machine (RF-SVM) stacking model exhibits the best predictive performance and enables the reliable determination of AA and ALP in real samples.