Information fusion via constrained principal component regression for robust quantification with incomplete...

Area of Science:

• Chemometrics
• Analytical Chemistry
• Spectroscopy

Background:

• Incomplete calibrations hinder chemometric data analysis, especially with complex matrices or unpredictable sample compositions.
• Existing regression models prioritize data explanation over chemical meaningfulness, leading to inaccurate results.

Purpose of the Study:

• To develop a novel chemometric approach for accurate quantitative analysis using experimentally feasible external calibrations.
• To improve the reliability and reproducibility of chemical analysis in complex biological samples.

Main Methods:

• Developed an 'ex situ calibration' method using analyte concentration series outside the chemical matrix.
• Incorporated external knowledge as regression constraints (e.g., literature values, stoichiometry) to compensate for missing information.
• Applied principal component regression (PCR) with constraints for quantitative analysis of compounds in microalgae cells using FTIR spectra.

Main Results:

• An incomplete ex situ calibration model initially failed for microalgae analysis due to matrix complexity.
• Incorporating regression constraints into PCR predictions successfully avoided chemically impossible results.
• Concentration reproducibility was significantly enhanced, with errorbars reduced by an order of magnitude for most samples.

Conclusions:

• The novel chemometric method effectively overcomes limitations of incomplete calibrations in complex matrices.
• This approach enhances the accuracy and reproducibility of quantitative analyses, enabling detailed studies of cellular responses to environmental chemical shifts.