Figure 1.

Flow chart of the data analysis. The figure provides an overview on the applied machine-learning approach, which was performed in three main steps (left column of boxes, indicated in blue letters): Firstly, data preprocessing was applied as indicated in the grey boxes. This provided lipid markers of three different classes (white rectangles) that formed the data spaces. Secondly, structure identification in the data space was performed by means of unsupervised machine-learning with the emergent self-organizing feature maps (ESOM)/U-matrix method [12]. Thirdly, the identified structures were superimposed with the known grouping of the data into multiple sclerosis (MS) patients and healthy subjects. The agreement between identified and known structure was then analyzed (structure interpretation).