We propose a new lipidomics data normalization strategy that consist of normalizing each lipid species data by the sum of its class, followed by a centered log-ratio transformation (clr [1]). This normalization approach was compared with log2 approach classically used for semi-targeted lipidomics data obtained by supercritical fluid chromatography-mass spectrometry [2], on milk collected from dairy ewes. The final aim of this study is to identify non-invasive biomarkers of lipolysis in ewe's milk.

The rapid development of mass spectrometry-based lipidomics has enabled the simultaneous measurement of a large number of lipid species from different classes. As a new field, the analysis of semi-targeted lipidomic data has been strongly influenced by metabolomic methodologies. The most common strategy is to apply a logarithmic transformation. We propose to normalize each lipid by the sum of its class, followed by a clr transformation to account for the compositional aspect.

A lipidomic dataset obtained from dairy ewes was normalized by clr or by the classical log2 transformation. The impact of these normalization methods for discovering milk lipolysis biomarkers was then investigated, focusing on the interpretability of the results. Graphical evaluations showed animals were more comparable with the clr approach. Differential analyses found 14 significant lipids for the log2 and 19 for clr with only 6 in common, and are currently being interpreted biologically.