spatialstein: An Open-Source Workflow for Annotation, Deconvolution, and Spatially Aware Segmentation of Mass Spectrometry Imaging Data

Abstract

Mass Spectrometry Imaging (MSI) data sets are markedly different from optical images. However, analysis algorithms often overlook the intricacies of this kind of data. In MSI, a frequently observed phenomenon is variability in signal intensity between pixels caused by factors other than differences in analyte concentrations. Another common issue is the presence of ions with overlapping isotopic envelopes resulting in isobaric interference. The first factor causes random variations of the signal from the same anatomical regions. The second can cause the spatial distribution of a single peak to represent a mixture of spatial distributions of several analytes. Both factors affect the accuracy of data analysis methods such as MSI segmentation. In this article, we demonstrate that accounting for the intricate structure of MSI data can increase the accuracy of the analysis results. We propose an approach that leverages recent advancements in computational mass spectrometry to separate overlapping isotopic envelopes and mitigate pixel-to-pixel variability of signal intensity. We implemented the approach in spatialstein, an open-source workflow that provides a tentative annotation of an MSI data set with molecular formulas, generates a deconvolved ion image for each annotated ion, and segments each deconvolved ion image into regions of distinct intensity of the corresponding analyte. The structure of the workflow is modular, making it highly modifiable and applicable, whole or in parts, to other studies. The spatialstein workflow is available at https://github.com/mciach/spatialstein.