Study On Antibiotics Raman Spectroscopy Based On Partial Least Squares

With the prominence of ecosystem pollution problem caused by antibiotics, thedetection of antibiotics has become an issue of increasing concern. Since thetraditional detection methods are expensive and complex to operate, it’s necessary tofind a low cost, fast and suitable method for the detection of various antibiotics.This paper introduced the fundamental theory of qualitative discrimination andquantitative analysis based on Partial Least Squares (PLS), the fundamental theory,advantage and analysis process of the Raman spectra, and also introduced theapplication of Raman spectra in antibiotics. The antibiotics used in our paper containsDoxycycline (DOTC, tetracyclines), oxytetracycline (OTC, tetracyclines), and tylosin(TYL, macrolides). Antibiotics identification and aqueous mixture quantification wasrealized by the combination of PLS and portable Raman spectroscopy. The researchcontains the establishment of the calibration set and the elimination of the singularsample, the impact of variable selection based on PLS projection analysis anddifferent pretreatment on the model, the threshold optimization of the discriminantmodel, and so on.The aqueous solutions of doxycycline(DOTC), oxytetracycline(OTC), andtylosin(TYL) were analyzed by Raman spectroscopy respectively. The characteristicRaman peaks were extracted and assigned to the bands of Raman spectrum. Then thequalitative discriminant model was constructed based on the special spectral range.The model optimization was conducted by variable selection, spectral pretreatment,and threshold selection. Forecast accuracy of unknown samples with the optimizationmodel is come to100%.The Raman spectroscopy was used to analyze mixed solution of the antibiotics.The orthogonal experimental design was used to establish the calibration set andprediction set, and then the model was optimized and discussed from the eliminationof singular sample, spectra pretreatment and variable selection. To some extent, theprediction results were good using the optimization model to analyze the externalsamples.Result in our paper showed the methods described here were feasible for rapidand easy determination of antibiotic components in aqueous solution. The algorithms proposed in this paper obtained a good performance, in this case, quick and realizablequalitative and quantitative analysis could be conducted.