Exploring High-Dimensional LA-ICP-TOFMS Data with Uniform Manifold Approximation and Projection (UMAP)

Spectral imaging generates information-rich datasets comprising a large map of pixels that each contain a comprehensive spectrum. A specific form of mass spectral imaging is laser ablation-inductively coupled plasma-time-of-flight mass spectrometry (LA-ICP-TOFMS). This technique enables elemental imaging of almost the entire periodic table. The large number of isotopes per pixel leads to high-dimensional data posing major challenges for visualisation, pattern recognition and interpretation. To decrease this complexity, dimensionality reduction techniques, such as uniform manifold approximation and projection (UMAP), provide powerful tools to transform high-dimensional datasets into low-dimensional representations aiming to preserve data point relationships and visualise spectral similarities. This study provides a detailed introduction to UMAP for analysing LA-ICP-TOFMS data. By transforming high-dimensional MS imaging data into two-dimensional spaces, UMAP facilitates automated visualisation to identify spectral clusters. UMAP’s utility to reveal spectrally distinct regions and tissue heterogeneity is demonstrated for a chicken embryo and a honeybee specimen. For detailed cluster analysis, a hierarchical strategy is introduced involving iterative UMAP applications, first to the global dataset, and then to resulting clusters. This approach helps uncover subtle chemical patterns hidden in the initial global UMAP application. Furthermore, the influence of the most relevant UMAP hyper-parameters is discussed, providing guidance for selecting critical parameters for further datasets. Overall, this study introduces UMAP as an exploratory and versatile tool for targeted and non-targeted analysis of complex LA-ICP-TOFMS data. Its integration into imaging workflows supports spectral clustering, image segmentation, hypothesis generation, and rapid analysis of large and high-dimensional spectral data from biological and environmental specimens.