Detection and visualization of communities in mass spectrometry imaging data

Area of Science:

• Metabolomics
• Bioinformatics
• Computational Biology

Background:

• Investigating spatial distribution and colocalization of metabolites is crucial for understanding molecular network functions.
• Existing methods may not fully capture the complex spatial relationships and functional implications of metabolite distributions.
• Community detection offers a novel approach to group molecules with correlated spatial patterns, suggesting functional associations.

Purpose of the Study:

• To develop and apply a community detection method for analyzing mass spectrometry (m/z)-images of metabolites.
• To group molecules with similar spatial distributions, inferring potential functional networks or pathway activities.
• To provide an interactive visualization tool for exploring these molecular communities.

Main Methods:

• Utilized a spectral approach for community detection by optimizing the modularity measure on m/z-images.
• Developed an analysis pipeline integrating community detection with spatial data.
• Created an interactive web-based visualization tool (GRINE) for explorative data analysis.

Main Results:

• Successfully identified molecular communities with correlated spatial distributions in barley seed and glioblastoma datasets.
• Reproduces previous findings on barley seed anatomy and confirms localized molecular compositions in glioblastoma, aligning with histology.
• Discovered novel subcommunities within m/z-images, revealing finer-scale metabolite distribution patterns and enabling detailed investigation via the GRINE tool.

Conclusions:

• The proposed community detection method effectively identifies groups of laterally co-localized molecules.
• Detected molecular communities exhibit substructures, readily explorable with the interactive visualization tool.
• This approach serves as a valuable complement to existing pixel clustering methods for spatial metabolomics data analysis.