Clustering-based preprocessing method for lipidomic data analysis: application for the evolution of newborn skin surface lipids from birth until 6months.

Abstract

After life in utero and birth, the skin is submitted to an important process of adaptation to a relatively dry gaseous environment. Skin surface lipids (SSLs) contribute actively to the protection of the skin barrier. Within this context, our objective was to study the evolution of each lipid compound during the postnatal period. SSLs were collected from six newborns a few days after birth until the age of 6months. Seventy samples were analyzed using high-temperature gas chromatography coupled to mass spectrometry (HT-GC/MS). The use of separative techniques coupled to mass spectrometry for the analysis of samples containing complex mixtures of lipids generates a large volume of data which renders the results interpretation very difficult. In this study, we propose a new approach to handle the raw data, a clustering-based preprocessing method (CB-PPM), in order to achieve (1) volume reduction of data provided by each chromatogram without loss of information, (2) alignment of time retention shift between different runs, (3) clustering of mass spectra of the same molecule in one qualitative group, (4) and integration of all data into a single matrix to be explored by chemometric tools. This approach allowed us to gather data variations in 256 qualitative groups and therefore enabled us to highlight the variation of compounds including those of low intensity. Moreover, the representation of all data gathered in one matrix rendered reading of the results rapid and efficient. Thus, using this approach, we have demonstrated an increase of cholesterol esterification with epidermal fatty acids (C20 to C25) with age. This epidermis participation in SSL production at a molecular level in the first period of life has not been previously shown. These data can be very interesting for the development and improvement of products destined for the protection of infant skin. Graphical abstract ᅟ.