Pharmaceutical Analysis Model Robustness From Bagging-PLS and PLS Using Systematic Tracking Mapping.

Area of Science:

• Analytical Chemistry
• Chemometrics
• Spectroscopy

Background:

• Quantitative modeling is crucial for analyzing complex datasets.
• Traditional step-by-step optimization methods can be suboptimal for model development.
• Near-infrared (NIR) spectroscopy offers a non-destructive analytical technique.

Purpose of the Study:

• To investigate the efficacy of a processing trajectory approach for optimizing multivariate calibration models.
• To compare the performance of processing trajectory with step-by-step optimization.
• To enhance the accuracy and robustness of quantitative models using advanced variable selection and ensemble methods.

Main Methods:

• Utilized interval partial least-squares (iPLS), backward interval partial least-square (BiPLS), and synergy interval partial least-squares (SiPLS) for variable selection.
• Developed quantitative models using Partial Least Squares (PLS) and Bagging-PLS.
• Assessed model performance using root mean square errors of validation (RMSEP) and ratio of standard error of prediction to standard deviation (RPD).

Main Results:

• The processing trajectory method yielded more reliable and robust quantitative models compared to step-by-step optimization.
• Variable selection techniques significantly improved prediction accuracy.
• Bagging-PLS models demonstrated enhanced prediction accuracy and robustness.

Conclusions:

• The processing trajectory approach is advantageous and feasible for developing and optimizing multivariate calibration models.
• Systematic tracking effectively optimizes modeling parameters, leading to superior model performance.
• The combination of advanced variable selection and ensemble methods (Bagging-PLS) improves prediction accuracy and robustness in NIR analysis.