Interpretation of matrix chromatographic-spectral data modeling with parallel factor analysis 2 and multivariate...

Area of Science:

• Chemometrics
• Analytical Chemistry
• Spectroscopy

Background:

• Parallel factor analysis 2 (PARAFAC2) is commonly used for second-order chromatographic-spectral data analysis.
• Existing PARAFAC2 models, including the non-negative NN-PARAFAC2, impose constraints that limit their general applicability.
• These constraints can lead to significant bias in quantitative analysis of analyte concentrations.

Purpose of the Study:

• To investigate the limitations of PARAFAC2 and NN-PARAFAC2 in modeling chromatographic-spectral data.
• To compare the performance of PARAFAC2 methods with multivariate curve resolution-alternating least-squares (MCR-ALS).
• To evaluate the quantitative accuracy for determining a fluoroquinolone antibiotic in bovine liver samples.

Main Methods:

• Simulations were used to understand and visualize PARAFAC2 model behavior.
• Experimental data analysis involved liquid chromatography with multi-wavelength fluorescence detection.
• Comparison of PARAFAC2, NN-PARAFAC2, and MCR-ALS for data processing and analytical performance.

Main Results:

• Both PARAFAC2 and NN-PARAFAC2 showed poor analytical results on simulated and experimental data.
• MCR-ALS provided accurate data processing and reliable analytical indicators, outperforming PARAFAC2 methods.
• MCR-ALS achieved significantly lower errors and better recoveries compared to PARAFAC2 for quantitative determination.

Conclusions:

• The inherent constraints in PARAFAC2 and NN-PARAFAC2 limit their effectiveness for general chromatographic-spectral data.
• MCR-ALS is a more robust and accurate method for quantitative analysis in complex matrices.
• MCR-ALS demonstrates superior performance in terms of error reduction and recovery accuracy.