PdO-T nanozyme-based three-channel colorimetric sensor array coupled with machine learning for accurate differentiation and quantitative detection of aflatoxins in U.S. pecan.

Aflatoxins (AFs) represent a class of highly toxic food contaminants, yet existing detection approaches still suffer from several critical limitations. Herein, three PdO nanoenzymes (denoted as PdO-160, PdO-200, and PdO-240) were synthesized by regulating the hydrothermal temperature. All three nanozymes display robust and stable peroxidase-like activity via the generation of •OH and 1O₂. By exploiting the distinct inhibitory effects of AFB1, AFB2, AFG1, and AFG2 toward these nanoenzymes, a three-channel colorimetric sensor array was rationally constructed. Combined with a random forest (RF) model, the proposed array enables accurate discrimination of all four AFs independent of concentration. Further quantitative prediction was conducted using RF regression and artificial neural network (ANN) models, and the RF regression delivers optimal fitting performance with the maximum determination coefficient R2 up to 0.971 for the four target analytes. In real-sample analysis, the array obtains favorable classification performance for mixed aflatoxins, with an accuracy of 95.65% for concentration-independent identification, and all R2 values above 0.956 in quantitative analysis. The nanoenzymes also exhibit excellent batch-to-batch reproducibility and long-term storage stability. This study offers a promising colorimetric sensing strategy for the screening of aflatoxins in food matrices.