Probabilistic classification method on multi wavelength chromatographic data for photosynthetic pigments identification

• Main Authors: Prilianti, K R, Setiawan, Y, Indriatmoko, Adhiwibawa, M A, Limantara, L, Brotosudarmo T H P
• Format: Conference Proceeding
• Language: English
• Subjects: Carotene; Chromatography; Classification; Data processing; Fingerprinting; High performance liquid chromatography; Pattern recognition; Photosynthesis; Pigments; Principal components analysis; Probabilistic methods
• ISSN: 0094-243X; 1551-7616
• DOI: 10.1063/1.4866538

Environmental and health problem caused by artificial colorant encourages the increasing usage of natural colorant nowadays. Natural colorant refers to the colorant that is derivate from living organism or minerals. Extensive research topic has been done to exploit these colorant, but recent data shows that only 0.5% of the wide range of plant pigments in the earth has been exhaustively used. Hence development of the pigment characterization technique is an important consideration. High-performance liquid chromatography (HPLC) is a widely used technique to separate pigments in a mixture and identify it. In former HPLC fingerprinting, pigment characterization was based on a single chromatogram from a fixed wavelength (one dimensional) and discard the information contained at other wavelength. Therefore, two dimensional fingerprints have been proposed to use more chromatographic information. Unfortunately this method leads to the data processing problem due to the size of its data matrix. The other common problem in the chromatogram analysis is the subjectivity of the researcher in recognizing the chromatogram pattern. In this research an automated analysis method of the multi wavelength chromatographic data was proposed. Principal component analysis (PCA) was used to compress the data matrix and Maximum Likelihood (ML) classification was applied to identify the chromatogram pattern of the existing pigments in a mixture. Three photosynthetic pigments were selected to show the proposed method. Those pigments are β-carotene, fucoxanthin and zeaxanthin. The result suggests that the method could well inform the existence of the pigments in a particular mixture. A simple computer application was also developed to facilitate real time analysis. Input of the application is multi wavelength chromatographic data matrix and the output is information about the existence of the three pigments.