Application of quantitative chemometric analysis techniques to direct sampling mass spectrometry

Area of Science:

• Analytical Chemistry
• Chemometrics
• Environmental Science

Background:

• Analyzing complex mixtures with similar mass spectra poses challenges for traditional methods.
• Direct sampling mass spectrometry offers a direct approach but requires advanced data processing for mixtures.
• Multivariate chemometric techniques are essential for deconvoluting complex spectral data.

Purpose of the Study:

• To evaluate multivariate calibration models for analyzing water samples containing mixtures of toluene, ethyl benzene, and cumene.
• To compare the performance of partial least-squares regression (PLS), trilinear partial least-squares regression (tri-PLS), and parallel factor analysis (PARAFAC).
• To assess the superiority of multivariate methods over univariate regression, especially when unique ions are unavailable for quantification.

Main Methods:

• Purge-and-trap/direct sampling mass spectrometry was employed for sample analysis.
• Multivariate calibration models including PLS, tri-PLS, and PARAFAC were constructed.
• PARAFAC and tri-PLS leveraged analyte temporal profiles for improved analysis.
• Model performance was evaluated against univariate regression using prediction errors.

Main Results:

• Multivariate quantitative methods demonstrated superior performance compared to univariate regression when unique ions were absent.
• PARAFAC exhibited enhanced accuracy in the presence of unmodeled, interfering compounds.
• The inherent uniqueness of PARAFAC facilitated the estimation and identification of interfering compound mass spectra.

Conclusions:

• Direct sampling mass spectrometry combined with multivariate analysis provides robust quantification of challenging mixtures.
• PARAFAC is a powerful tool for analyzing complex environmental samples, offering advantages in handling interferences.
• The ability of PARAFAC to estimate and identify interfering compounds enhances analytical confidence and sample characterization.