Conv2dPA: Two-dimensional Convolution Sharpening-Based Nonlinear Alignment Solution for Chromatographic Fingerprints with High Complexity and Co-elution Behavior

Retention time shifts across chromatographic fingerprints can negatively impact the reliability of quantitative and similarity analysis results. Hence, we present a novel two-dimensional convolution sharpening-based solution (Conv2dPA) for nonlinear peak alignment of co-eluting components in highly complex samples. The core of Conc2dPA is the exploitation of intrinsic bilinear structure in chromatographic-spectral data and the specially designed two-dimensional sharpening filters to compensate for the lack of chromatographic resolution and spectral selectivity between co-eluting analytes. After convolution, the chromatographic peak position of each analyte can be highlighted and accurately localized. Notably, Conv2dPA performs peak alignment through an iterative optimization procedure to improve the alignment accuracy of low-response or embedded peaks. In the final step, fingerprints can be reconstructed in batches with aligned chromatographic profiles and their corresponding spectra for subsequent analysis. Investigations on simulated datasets demonstrated that Conv2dPA has achieved proper alignment even in tricky situations of embedded overlap, poor spectral selectivity, unknown interferences, and variable baseline drift. Furthermore, in two experimental case studies, Conv2dPA has effectively attenuated the negative impact of nonlinear retention time shifts and successfully improved the reliability of quantification and similarity analysis, respectively.