Synchronous fluorescence spectra-based machine learning algorithm with quick and easy accessibility for simultaneous quantification of polycyclic aromatic hydrocarbons in edible oils

Polycyclic aromatic hydrocarbons (PAHs) are one of the leading causes of human cancer. Four typical PAHs (PAH4) including benzo(a)pyrene (BaP), benzo(b)fluoranthene (BbF), benzo(a)anthracene (BaA), and chrysene (Chr) have been regarded as reasonable indicators for the occurrence of PAHs in food. In this study, the constant wavelength synchronous fluorescence (CWSF) spectra of PAH4 mixtures were used as the data sets without preprocessing and directly combined with the back propagation neural network (BPNN) algorithm to establish a quantitative analysis method of PAH4. This method is capable of predicting the concentrations of PAH4 in edible oil samples without pre-separation. The detection limits for BaP, BbF, BaA, and Chr were 0.014, 0.068, 0.026, and 0.013 mu g/kg, respectively. The recoveries in various oil samples for BaP, BbF, BaA, and Chr were 99.5 +/- 2.1, 101.0 +/- 4.6, 98.6 +/- 3.2, and 98.5 +/- 4.9 %, respectively. The proposed method has proved to be a powerful tool for the rapid detection of PAH4.

Automated summary

Polycyclic aromatic hydrocarbons (PAHs) are a major cause of human cancer. This study developed a quantitative analysis method for PAH4 using the back propagation neural network (BPNN) algorithm and constant wavelength synchronous fluorescence (CWSF) spectra as the data sets. The method can predict the concentrations of PAH4 in edible oil samples without preprocessing or pre-separation. It has been proven to be a powerful tool for the rapid detection of PAH4.