Interference-free calibration with first-order instrumental data and multivariate curve resolution. When and why?

Area of Science:

• Analytical Chemistry
• Chemometrics

Background:

• Multivariate curve resolution-alternating least-squares (MCR-ALS) is explored for first-order calibration.
• MCR-ALS can extract meaningful data even with unexpected species, unlike classical models like partial least-squares regression (PLS).
• Rotational ambiguity (RA) can negatively impact MCR-ALS predictive capacity in first-order datasets with interferents.

Purpose of the Study:

• To investigate the conditions and reasons for successful interference-free calibration using MCR-ALS with first-order data.
• To analyze the influence of signal overlapping, RA extent, and alternating least squares (ALS) initialization on MCR-ALS model performance.
• To alert analytical chemists about potential limitations of MCR-ALS for interference-free first-order calibration.

Main Methods:

• Simulated and experimental data were used to model a single-analyte, single-interferent calibration scenario.
• Analysis focused on signal overlapping, rotational ambiguity, and alternating least squares (ALS) initialization.
• The predictive performance of MCR-ALS models was evaluated under various conditions.

Main Results:

• MCR-ALS can achieve interference-free calibration with first-order data under specific conditions.
• Rotational ambiguity and the choice of ALS initialization significantly affect model accuracy.
• The success of interference-free MCR-ALS is not guaranteed and depends on data characteristics and modeling choices.

Conclusions:

• While MCR-ALS offers advantages for first-order calibration, its effectiveness in achieving interference-free results is contingent on factors like signal overlap and RA.
• Proper initialization of the alternating least squares (ALS) algorithm is crucial for optimizing MCR-ALS predictive performance.
• Analytical chemists should exercise caution and thoroughly evaluate MCR-ALS models to avoid potential pitfalls in interference-free calibration scenarios.