Large amount of data are generally generated from mass spectrometry imaging (MSI) experiments, and they contain very rich chemical and spatial information of substances in biological samples. However, large size and complex structure of MSI datasets make it challenging to deeply understand the overall metabolic information present in mass spectrometry (MS) images. Traditional MSI data analyses are generally carried out using manually selected ions. Effective data analysis methods need to be developed to efficiently extract crucial information from large amounts of MSI data. The Single-probe, a microscale sampling and ionization device, has been coupled to mass spectrometer to generate high-spatial resolution MS images of mouse kidney slice under ambient conditions. We then conducted advanced metabolic analysis of MSI data using multivariate curve resolution (MCR) and machine learning (ML) approaches. MCR was used to group molecules with similar patterns of spatial distribution. We have successfully constructed three major patterns with better clarified boundaries of histological structures in the tissue sample. In ML analyses, both supervised (i.e., random forest) and unsupervised (i.e., DBSCAN and CLARA) were applied to process the MSI data. Compared with MCR approach, the unsupervised ML discovered sub-regions with subtle features within the major regions. Although unsupervised ML methods provided features generally similar to those obtained from MCR and supervised ML approaches, they are relatively more efficient for the analysis of large amount of MSI data.

Combining the single-probe mass spectrometry imaging with advanced data analysis

255th National Meeting (2018)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.