Comparative analysis of peak-detection techniques for comprehensive two-dimensional chromatography

Significance of the topic

Comprehensive two-dimensional gas chromatography (GC×GC) offers unmatched resolving power for complex mixtures but demands robust peak-detection approaches to accurately quantify analytes, especially when rapid retention-time shifts occur. Reliable algorithms are essential for environmental monitoring, petrochemical analysis, food safety, and pharmaceutical quality control.

Objectives and overview

This study re-evaluates two widely used GC×GC peak-detection techniques — the two-step algorithm and the watershed algorithm — under unbiased conditions by applying retention-time shift correction to both. Through extensive simulations, the work aims to determine which method more accurately recovers peak volumes across varying noise levels, peak shapes, and chromatographic shifts.

Methodology and instrumentation

• Data simulation: Single 2D Gaussian peaks were generated with adjustable apex widths (σx, σy), added Gaussian noise (σn), and parametric skew to represent second-dimension retention-time shifts.
• Shift correction: Cross-correlation aligned secondary chromatograms prior to detection.
• Two-step algorithm: 1D detection in each secondary slice followed by retention-time overlap and unimodality merging.
• Watershed algorithm: 2D neighborhood growth from the global apex, assigning each point to existing peak labels based on intensity-ordered neighbors.
• Performance metrics: Repeated tests (n=1000) measured mean volume, standard deviation, error relative to true volume, and failure rate when detected peaks deviated beyond one standard deviation of true apex regions.

Results and discussion

• Both methods underestimate peak volumes as noise and peak width increase due to over-segmentation, but the watershed algorithm consistently yields smaller bias and higher accuracy.
• Under low noise, both perform similarly, but at higher σn (≥0.005) and broader peaks (σy≥4), watershed exhibits up to 4× lower volume error and far fewer detection failures (39 vs. 341 per 1000 trials).
• Statistical tests confirm the superiority of watershed detection (p<0.001) for most parameter combinations.
• Visualization illustrates that the two-step approach can split a single peak when local slice noise breaks continuity, whereas watershed’s 2D neighborhood analysis preserves peak integrity.

Benefits and practical applications

• Enhanced quantitative accuracy in GC×GC profiling of environmental samples, petrochemical streams, and biological fluids.
• Reduced false splitting of peaks improves compound identification and integration in automated workflows.
• Compatibility with existing GC×GC data processing pipelines when shift correction is implemented uniformly.

Future trends and potential applications

• Integration of advanced denoising or smoothing filters prior to peak detection to further suppress over-segmentation.
• Extension to multivariate and spectral deconvolution methods for resolving coeluted peaks.
• Real-time implementation in software controlling GC×GC platforms to adaptively correct shifts and detect peaks on the fly.
• Application to liquid chromatography×LC×LC and emerging two-dimensional separations in metabolomics and proteomics.

Conclusion

When retention-time shifts in the second dimension are corrected for both methods, the watershed algorithm demonstrates superior accuracy and robustness in detecting resolved GC×GC peaks across a broad range of noise levels and peak widths compared to the two-step approach. Adoption of watershed-based detection with proper shift alignment can significantly enhance the reliability of comprehensive 2D chromatographic analyses.

References

1. Beens J., Boelens H., Tijssen R., Blomberg J. High Resolut. Chromatogr. 21 (1998) 47.
2. Song Q., Savant A., Reichenbach S. SPIE Proc., Digital Image Processing (1999) p.2.
3. Reichenbach S., Ni M., Kottapalli V.V.A. Chemom. Intell. Lab. Syst. 71 (2004) 107.
4. Vivó-Truyols G., Janssen H.-G. J. Chromatogr. A 1217 (2010) 1375.
5. Ramos L., Sanz J. In: Comprehensive Two-dimensional Gas Chromatography, Elsevier (2009) p.283.
6. Skov T., Hoggard J.C., Bro R., Synovec R.E. J. Chromatogr. A 1216 (2009) 4020.
7. Peters S., Vivó-Truyols G., Marriott P., Schoenmakers P. J. Chromatogr. A 1156 (2007) 14.
8. van der Klift E.J.C. et al. J. Chromatogr. A 1178 (2008) 43.
9. Beucher S. Scanning Microsc. Int. 6 (1992) 299.
10. Reichenbach S. In: Comprehensive Two-dimensional Gas Chromatography, Elsevier (2009) p.77.
11. Bruckner C.A., Prazen B.J., Synovec R.E. Anal. Chem. 70 (1998) 2796.
12. Prazen B.J. et al. J. High Resolut. Chromatogr. 23 (2000) 215.
13. Fraga C.G., Prazen B.J., Synovec R.E. Anal. Chem. 73 (2001) 675.
14. Prazen B.J. et al. Anal. Chem. 73 (2001) 5677.
15. Sinha A.E. et al. J. Chromatogr. A 1027 (2004) 269.
16. Sinha A.E. et al. J. Chromatogr. A 1056 (2004) 145.
17. Fraga C.G., Corley C.A. J. Chromatogr. A 1096 (2005) 40.
18. Kong C., Ye F., Lu X., Guo L., Tian J., Xu G. J. Chromatogr. A 1086 (2005) 160.
19. Kong H., Ye F., Hugel H.M., Synovec R.E. Anal. Chem. 79 (2007) 1611.
20. Zeng Z.-D., Chin S.-T., Marriott P.J. J. Chromatogr. A 1218 (2011) 2301.