A Sensitive and Rapid Detection Method for the Stx2 Gene of Escherichia coli O157:H7, based on the Combination of Colorimetric Loop-Mediated Isothermal Amplification (cLAMP) with CNN Deep Learning Model

BACKGROUND: Escherichia coli producing Shiga toxin has been reported as a major problem in foodborne and clinical settings, primarily because the organism produces Shiga toxin type 2 (stx2). Therefore, this study aimed to establish a sensitive, rapid detection method for the stx2 gene of E. coli O157:H7 by combining a cLAMP method with a convolutional neural network (CNN) deep learning model. RESULT: The cLAMP conditions were optimized by varying the incubation temperature and reaction time. A positive reaction led to a visible color change from pink to yellow. Subsequently, the assay was evaluated for specificity and sensitivity. The optimal cLAMP incubation time and temperature were 60°C for 25 minutes. The method indicated high sensitivity and specificity, with an LOD of 14.7 fg. The cLAMP was combined with CNN deep learning. The model was trained through transfer learning and assessed on the test dataset based on precision, accuracy, F1-score, and recall. The results showed that a rapid and sensitive cLAMP assay combined with CNN deep learning was successfully developed. CONCLUSION: A combination of cLAMP and CNN using the MobileNetV2 architecture offers the highest specificity, sensitivity, and accuracy among the evaluated architectures. This combination has the potential for point-of-care testing (POCT).