Random forest algorithm-enhanced dual-emission molecularly imprinted fluorescence sensing method for rapid detection of pretilachlor in fish and water samples

Abstract

A sensitive and efficient fluorescence sensor based on dual-emission molecularly imprinted polymers (Dual-em-MIPs) was successfully developed using the random forest (RF) machine-learning algorithm for the rapid detection of pretilachlor. SiO2 coatings on red-emitting CdSe/ZnS quantum dots (r-SiO2@QDs) as intermediate light-emitting components are non-selective for pretilachlor, whereas molecularly imprinted layers coated with blue-emitting nitrogen-doped graphene quantum dots (N-GQDS) are selective. Fluorescence images of the Dual-em-MIPs were acquired. The red (R), green (G), and blue (B) color values of the image were analyzed using an RF algorithm, and the classifier was trained using 103 fluorescent images for automatic analyses. Under optimized conditions, an excellent linear relationship between the sensor and pretilachlor was obtained in the range of 0.001–5.0 mg/L (R2, 0.9958). Additionally, the satisfactory recoveries of Dual-em-MIPs ranged between 92.2 % and 107.6 % for the real samples, with a relative standard deviation (RSD) under 6.5 %. The satisfactory recoveries of the RF model based on the fluorescence sensor were 84.2–108.2 % with the RSD under 6.4 %. Overall, the proposed fluorescence sensor based on Dual-em-MIPs and machine learning methods was successfully used to determine pretilachlor in the environment and in aquatic products.