Analyzing imaging mass spectrometry data by implementation of sparsity Constraint on multivariate curve resolution

In this work, the potential of data analysis methods in obtaining quantitative and qualitative chemical information from mass spectrometry imaging (MSI) data about biological tissue will be studied. In recent years, due to the increasing high-resolution power of mass spectrometers, the number of the mass-to-charge (m/z) of mass spectra has also become very high, which leads to the production of very large data [1]. So, in raw data, the amount of information obtained from the sum of all spectra is very more than the required value. Ignoring much of the data does not fully exploit the potential of MSI techniques [2]. Using this method, both spatial distribution and spectral information of analyzed samples can be obtained. However, these main MSI bugs are mostly dependent on the complexity of the real target samples and the huge size of the data generated by this method [3]. Therefore, chemometrical tools are used to solve this problem. The present contribution is about the implementation of sparsity constraint in multivariate curve resolution-alternating least square (MCR-ALS) technique for analysis of three-way MSI data arrays (two spatial and one m/z dimensions) [4]. In the present study, the L1-regularization paradigm has been implemented in each iteration of the MCR-ALS algorithm in order to force the algorithm to return sparser mass spectra. L1-regularization has been applied by using the least absolute shrinkage and selection operator (Lasso) instead of the ordinary least square [5]. The advantages of the application of this method are shown for several examples of different data consisting of two three-way simulated data with 3 components that each component has a specific pattern to examine its effect on their interactions with each other and the background and also on a real MS data image of a mouse brain thin layer that was obtained by MALDI-MSI at (NERSC) [6]. The simulated data in this work have been fabricated and studied in different background modes with and without overlapping m/z channels, which makes this study comprehensive. Finally, the effect of the sparsity constraint has been explored on the possible solutions in MCR methods is investigated. The results in this work showed that the implementation of this constraint reduces the extent of rotational ambiguity in MCR results and can be useful for solving MS data with overlapping in mass spectra and concentration profiles [4].