Development of optimized regression model for Raman spectral analysis of pharmaceutical drugs cefixime and sitagliptin

To build a robust model for the quantification of pharmaceutical drugs, different regression algorithms and preprocessing techniques are employed. In this study, Raman spectroscopy is used for the analysis of solid dosage formulations of two drugs, cefixime and sitagliptin. Using the different amounts of excipients and active pure ingredients (API), the different concentrations of cefixime and sitagliptin are made. The principal component regression (PCA), partial least-squares regression (PLSR), and PLSR based on feature selection by variable importance in projection (VIP) scores are used to regress the model to quantify the API concentration of cefixime and sitagliptin in a mixture of excipients. These models are built on both normalized and unnormalized data of cefixime and sitagliptin to check the performance of the constructed models. By comparing errors of all trained models, it is observed that the root mean square error of calibration (RMSEC) and root mean square error of prediction (RMSEP) were minimum in the case of PLSR model construction on unnormalized data of both cefixime and sitagliptin.

Automated summary

This study analyzes solid dosage formulations of two drugs using Raman spectroscopy and builds a quantification model using different regression algorithms and preprocessing techniques. The performance of the models constructed on normalized and unnormalized data is compared, and it is found that the PLSR model constructed on unnormalized data has the minimum error.