|Name||Mr. Nicholas Ellin|
|Organization or Institution||University of Florida|
Extended similarity methods for efficient data mining in imaging mass spectrometry
Nicholas R. Ellin,1 Colton M. Hunt, Boone M. Prentice1 and Ramón Alain Miranda-Quintana1
1 Department of Chemistry, University of Florida, Gainesville, FL, 32611-7200; USA.
Imaging mass spectrometry (IMS) is a label-free imaging modality that allows for the spatial mapping of many biological molecules directly in tissue. In an IMS experiment, a raster of the tissue surface produces a mass spectrum at each sampled position, resulting in thousands of individual spectra that form pixels in the final ion images. Each spectrum contains several thousand compounds at discrete m/z values that result in unique ion images. The high dimensionality of IMS data makes data processing and analysis difficult and time-consuming. Post-processing techniques, such as principal component analysis (PCA), have emerged as useful tools for mining IMS datasets to identify biological regions of interest and more thoroughly understand tissue biochemistry. One challenge with PCA in IMS is the interpretation of the loadings and scores. For example, the loadings often contain negative peaks in the PCA-derived pseudo-spectra, which are difficult to ascribe to the underlying biology. We have recently developed novel extended similarity indices, which allow us to more efficiently compare large volumes of IMS spectra simultaneously. In this method, each spectrum is represented as a binary fingerprint and the extended similarity indices tally the number of coinciding 1’s and 0’s to determine the physical similarity of all selected spectra.