Intelligent Peak and Spectrum Deconvolution Using Photodiode Array Detector

Significance of the Topic

Chromatographic peak overlap is a common challenge in analytical chemistry, complicating accurate quantitation and spectral interpretation. Modern detectors such as photodiode array (PDA) systems generate three-dimensional data combining retention time, wavelength, and absorbance, offering rich information for resolving co-eluting compounds. Intelligent deconvolution enhances data utility in quality control, pharmaceutical analysis, and environmental monitoring by extracting pure component profiles without complete physical separation.

Objectives and Study Overview

This work introduces i-PDeA II, an advanced peak and spectrum deconvolution algorithm designed to:

• Automatically resolve two- and three-component systems from PDA data.
• Assess deconvolution performance over varying component abundance ratios and chromatographic resolutions.
• Demonstrate practical applications in size-exclusion chromatography and ultra-fast HPLC.

Methodology and Instrumentation

The i-PDeA II approach integrates multivariate curve resolution–alternating least squares (MCR-ALS) with an expectation-maximization (EM) framework constrained by a bidirectional exponentially modified Gaussian (bemg) model. Key steps include:

1. Representing the three-dimensional PDA dataset D as the sum of outer products of concentration profiles (C) and spectra (S^T).
2. Applying MCR-ALS to minimize residual error, alternating between updating concentration estimates and spectral estimates.
3. Employing the bemg function to model chromatographic peak shapes, accounting for tailing and fronting.
4. Automatically optimizing the number of components.

Instrumentation used:

• HPLC system equipped with a Shimadzu photodiode array detector.
• Size-exclusion chromatography columns for polymer additive analysis.
• Data processing software implementing MCR-ALS and EM algorithms.

Main Results and Discussion

2-Component System:

• Cytidine and AMP were deconvoluted with area recovery errors below 1.8% and spectral similarity >0.9995.
• Relative standard deviations (n=6) remained under 3% for mixture analyses.

3-Component System (Abundance Ratio 100/100/1):

• o-, m-, and p-methoxyacetophenone yielded errors below 4% for the trace component and spectral similarity ≥0.9996.

Effect of Chromatographic Resolution:

• At resolution (Rs) 0.2, deconvolution errors exceeded 30–60%. Below Rs 0.6, errors dropped to under 1% for all components.

Practical Application 1 – Polymer Additives in SEC:

• Three stabilizers (Irganox 1010, Tinuvin 144, Tinuvin 120) were quantified over 0.01–0.1% (w/v) with calibration R²>0.995 and RSDs <2%.

Practical Application 2 – Ultra-Fast HPLC:

• Caffeine, ethylparaben, and acetophenone were separated within 0.025 min; deconvolution errors remained below 2%, outperforming simple vertical integration which showed up to 13% error.

Benefits and Practical Applications of the Method

• Enables accurate quantitation at any selected wavelength despite incomplete separation.
• Offers an alternative to mass spectrometry for isomeric mixtures with identical m/z values.
• Preserves reliable spectral data for component identification post-deconvolution.

Future Trends and Opportunities

• Integration of i-PDeA II with high-throughput and ultra-high-pressure liquid chromatography platforms for rapid screening.
• Extension to complex biological matrices and environmental samples with overlapping UV absorbance profiles.
• Incorporation of machine learning approaches to optimize component number selection and model parameters dynamically.

Conclusion

i-PDeA II represents a robust, automated solution for chromatographic peak deconvolution using PDA data. It delivers high accuracy and precision across diverse sample types and separation speeds, enhancing analytical workflows where complete chromatographic resolution is impractical.

Reference

S. Arase et al., Journal of Chromatography A 1469 (2016) 35