Antibiotics discrimination in terahertz spectral database

Abstract.

Antibiotics residues in food or agricultural products has been a risk on human health, so antibiotics detection and discrimination become very important. To discriminate different kinds of antibiotics, a terahertz spectral database, which contains the functions of spectral processing, spectral analysis and spectral matching, is necessary. In the terahertz(THz) region, characteristic absorption peaks of many molecules like antibiotics are not obvious, and spectral matching algorithm can help to find the similarity between two spectra, so the suitability of spectral algorithm is essential to discriminate antibiotics in terahertz spectral database. Jaccard similarity coefficient, spectral angle and Pearson correlation coefficient, are similarity calculations that were used widely in various research fields. In this paper, these three calculations were transformed and used as spectral matching algorithms. Two digitized spectra are compared directly by these algorithms, point by point, and the outcome being a parameter quantifies the degree of similarity between the spectra. Moreover, a THz spectral database system (SDBS) including original software, combining with spectral matching algorithms mentioned, was established and used to discriminate these antibiotics by their absorption spectra. According to the matching results, antibiotics discrimination can be realized by the three algorithms, which means the spectral matching algorithms are expected to be suitable for other substance discrimination besides antibiotics.