PGAlign: Peak Group Alignment of Mass Spectrometry Data

Modern imaging mass spectrometry (MSI) technologies, such as matrix-assisted laser desorption/ionization (MALDI), provide unprecedented spatial resolution for mapping tissue biomolecules. However, experimental variations frequently cause mass-to-charge (m/z) shifts across different data batches, misaligning identical biological substances. Existing alignment algorithms often fail to project matched spectra onto the exact same m/z coordinates as the reference data, hindering downstream comparative analyses. Here, we propose PGAlign (Peak-Group Alignment), a novel peak-group-based algorithm that aligns unknown spectra directly to a reference, producing matched data sets defined on a shared m/z grid. Operating on representative mean spectra, PGAlign executes four discrete steps: peak calling, peak grouping, peak-group pairing, and fine alignment assessment. Using simulated mouse pancreatic MALDI data and metabolomics data from a macrophage efferocytosis study, we show that PGAlign achieves lower alignment error than other methods. In canine sarcoma MALDI data, PGAlign preserves biologically meaningful differences between cancer and normal tissues and supports downstream classification, suggesting that alignment onto a common m/z grid can improve the reliability of MS-based biomarker analyses.