Variable clustering and spectral angle mapper-orthogonal projection method for Raman mapping of compound detection in tablets

Raman mapping and chemometrics are proposed to accurately characterize the composition of tablets. The most critical step of the state-of-art curve resolution methods (such as multivariate curve resolution-alternating least squares [MCR-ALS]) is the determination of the number of constituents, when chemical imaging is coupled with multivariate data analysis. However, it is usually performed in a considerably subjective way. We propose a variable clustering approach for the identification of the main dimensionality of vibrational spectral data. The method was tested on a Raman map of a complex pharmaceutical tablet that contained 4 major components with high spectral resemblance, and a low-dose lubricant was also added for tableting purposes. Using a variable clustering algorithm called VARCLUS we were able to construct clusters from the Raman mapping data corresponding to the real constitution of the sample. The modeled clusters were analyzed by the sum of ranking differences method. All 4 major components could be identified. The potential of the clustering algorithm was further assessed by applying MCR-ALS and spectral angle mapper-orthogonal projection methods. We have shown that variable clustering corresponded with MCR-ALS results and that it can be used to characterize the qualitative composition of an unknown pharmaceutical sample by combining the clustering algorithm with a pure component resolution method. Therefore, this method is well applicable to analyze and interpret the curve resolution of complex samples. Testing of the previously studied spectral angle mapper-orthogonal projection method, which relies on spectral reference libraries and even the low-dose lubricant (approximately 1% w/w), was identified through the chemical imaging. A variable clustering can be identified the main dimensionality of vibrational spectral data. The method was further assessed by applying multivariate curve resolution-alternating least squares (MCR-ALS) and spectral angle mapper-orthogonal projection methods. The variable clustering corresponded with MCR-ALS results, and it can be used to characterize the qualitative composition of an unknown pharmaceutical sample by combining the clustering algorithm with a pure component resolution method. Therefore, this method is well applicable to analyze and interpret the curve resolution of complex samples.