Integrative Clustering in Mass Spectrometry Imaging for Enhanced Patient Stratification

In biomedical research, mass spectrometry imaging (MSI) can obtain spatially-resolved molecular information from tissue sections. Especially matrix-assisted laser desorption/ionization (MALDI) MSI offers, depending on the type of matrix, the detection of a broad variety of molecules ranging from metabolites to proteins, thereby facilitating the collection of multilevel molecular data. Lately, integrative clustering techniques have been developed that make use of the complementary information of multilevel molecular data in order to better stratify patient cohorts, but which have not yet been applied in the field of MSI.

In this study, the potential of integrative clustering is investigated for multilevel molecular MSI data to subdivide cancer patients into different prognostic groups. Metabolomic and peptidomic data are obtained by MALDI-MSI from a tissue microarray containing material of 46 esophageal cancer patients. The integrative clustering methods Similarity Network Fusion, iCluster, and moCluster are applied and compared to non-integrated clustering.

The results show that the combination of multilevel molecular data increases the capability of integrative algorithms to detect patient subgroups with different clinical outcome, compared to the single level or concatenated data. This underlines the potential of multilevel molecular data from the same subject using MSI for subsequent integrative clustering.