New Approach for Chemometric Analysis of Mass Spectrometry Data

Abstract

The search of metabolites which are present in biological samples and the comparison between different samples allow the construction of certain biochemical patterns. The mass spectrometry (MS) methodology applied to the analysis of biological samples makes it possible for the identification of many metabolites. Each obtained signal (m/z) is characteristic of a particular metabolite. However, the mass data (m/z) interpretation is difficult because of the large amount of information that they contain. In this work, we present a relatively simple tool that allows us to deal with the whole of the mass information from the chemometric analysis. The statistical analysis is a key stage in order to identify the metabolites involved in a particular biochemical pattern. We transformed the mass data matrix in a vector. By having the data as a vector, it was possible to keep all the information and also avoid the signals overlapping, which is the major problem when the total ion chromatogram (TIC) is obtained. In the approach proposed here, the mass data (m/z) matrix was split in 100 different TIC in order to avoid the signal overlapping. The 100 chromatograms were concatenated in a vector. This vector, which can be plotted as a continuous (2D pseudospectrum), greatly simplifies for one to understand the subsequent dimensional multivariate analysis. To validate the method, 19 samples from two human embryos culture medium were analyzed by high-pressure liquid chromatography–mass spectrometry (HPLC–MS). Our methodology would be applied to the obtained raw data. Later on, a multivariate analysis was conducted using a robust principal components analysis interval (robPCA) and interval partial least squares algorithm (iPLS). The results obtained allow one to differentiate the two sample populations undoubtedly, although their composition was similar.