Abstract :
[en] 1 Introduction
Mass Spectrometry Imaging (MSI) is a powerful analytical tool allowing untargeted investigation of spatial distribution of molecular species in a large variety of samples, by recording 2D distribution of all the detectable compounds in the sample. Over the years, the application of MSI has become increasingly diverse, from bacteria-bacteria interaction understanding to biomarker discovery. To this end, it is necessary to combine high spectral resolution with efficient data processing, in order to maximise the extraction of the information contained in large datasets. Moreover, in addition to deal with thousands of features, pixel-to-pixel mass shift must be taken into account before applying any data-mining tool.
For that reason, we propose the combination of a post-acquisition label-free recalibration method and an algorithm to cluster features based on their Kendrick mass defect (KMD) in MSI.
2 Theory
KMD consists in a change of basis from the IUPAC mass scale to a Kendrick mass based on the nominal mass of a defined Kendrick mass unit, such as CH2 (here, nominal mass is set at 14 a.u.)1.
The KMD is obtained by subtracting the nominal Kendrick mass to the exact Kendrick mass. This mathematical transformation enables to map the detected molecules in mass spectrometry based on their chemical composition.
3 Material and methods
MALDIFT-ICR-MS (SolariX XR 9.4T) images were first converted into imzML open format. Each spectrum from the imzML file obtained for an MSI are individually peak picked. Then, our algorithm creates, for each pixel, a linear model linking m/z error and m/z. The model is creating by finding database hit with similar mass shift. Finally, a new imzML file is created by recalibrating each pixel with their estimated linear model.
Then, we developed a software to filter features (m/z peaks) based on their KMD from an imzML file. This software calculates first a KMD for each m/z values detected in the mean mass spectrum of the image. The MSI data is then filtered based on KMD value to conserve only ions whose KMD is included in the user-defined KMD range. The reduced ion lists were then divided into different compounds families, based on the repetition of the KMD. Finally, an image is generated for each compounds family. Thereby reducing the number of features of an image2.
4 Results and discussion
The KMD filtering method applied on a mouse brain tissue analysis by MSI appears to be biologically relevant since it enabled to map in a single step all members of important families of biomolecules. Among the detected families of biomolecules, lipids shows different distributions and localisations. Some of these lipids belonged to the glycerophosphocholines (GPCs), the hexosylceramides (HexCers), lysophoshocholins (LPCs) and the sphingomyelins (SMs) class. Moreover, The KMD analysis highlighted that some of the detected GPCs on the brain tissue sections from mouse were differentially co-localized, depending on their unsaturation degree. All these results suggested that the KMD could be considered instead of the mass-to-charge (m/z) values classically used for the analysis of images by MSI. The use of our in-house developed software for KMD analysis enables an automated, faster and efficient data analysis of MSI images.
5 Conclusion
The combination of recalibration before using Kendrick mass defect mass filter is an essential step to be able to interpret the image reconstruction of chemically-related compounds. This methods speed up the identification process and facilitates the data analysis without losing data information.
