Screening for dioxin contamination in fish oil by PARAFAC and N-PLSR analysis of fluorescence landscapes

A preliminary investigation of fish oils demonstrates that fluorescence excitation-emission landscapes evaluated by 3-way chemometric methods may be a candidate for an inexpensive screening method to indicate the level of contamination by dioxins and PCB’s which are normally analysed with expensive and time-consuming physicochemical separation techniques such as GC-MS.
Fluorescence landscapes of 88 fish oils have been investigated and showed great variation due to species, season and treatment, depicting a variation in natural fluorescent components. The fluorescence landscapes were analysed by PARAFAC. Samples with similar fluorescence fingerprints were selected from a PARAFAC score plot and local significant prediction models with PARAFAC/MLR, N-PLSR and PLSR were established with correlation coefficients in the range from r=0.69 (n=10) to r=0.97 (n=75) for dioxin and r=0.92 (n=12) for PCB. Application of PARAFAC/MLR and N-PLSR to fluorescence landscapes of fish oils resulted in local regression models for dioxin determination with prediction errors below 1 ng/kg, which is comparable to the reference method.
In the PARAFAC model, two of the modes give the excitation and emission spectra of the pure underlying fluorophores and the third mode their individual concentrations. Excitation and emission optima for 3-4 PARAFAC components in each data set were identified, representing both positive and negative (quenching) correlation components. It is hypothesized that the quenching correlation may be effected by the joint contribution of chlorinated organic compounds in the fish oil, including dioxins and PCB’s. Other explanations for the results are discussed.