Fig. 6: Real-time breath analysis offers a noninvasive window into altered metabolic pathways in patients not responding to pharmacotherapy and/or suffering from side effects.

a SESI–HRMS breath analysis detected alteration in the levels of several amino acids and associated compounds in epileptic patients. The figure shows a simple (unweighted, undirected, no loops, or multiple edges) graph of amino acid metabolism (based on KEGG map01230: biosynthesis of amino acids). Each node is a compound, where node-fill colour represents the mean log₂-scaled fold change in the side effects (yes vs. no), and drug-response (nonresponder vs. responders) dataset from the training set. Node-border colour represents whether the compound was assigned to significant or background list via MetaboAnalystR. Only compounds from the significantly enriched pathways (top-right quadrant in Fig. 5) are coloured. The rest are shown as grey. Node with two colours (node split) denotes that the compound was present under significantly enriched pathway of both datasets. b, c Density plot for the predicted score and classes of having side effects (b) and drug response (c). Density curves are accompanied with the actual data points, where each point represents one measurement from UKBB dataset, coloured based on clinically observed side effects (b) and clinically observed drug response (c). On predicted scores, a cutoff was assigned (based on Youden’s index calculated using only training-set data) to separate predicted classes.