Untargeted metabolite discovery in kinetic data from multi-dose intervention studies

Abstract

A new strategy for biomarker discovery is presented that is based on multi-dose kinetic metabolomics data. Gas chromatography-mass spectrometry (GC-MS) data sets recorded in the full scan mode are scanned for compounds showing a meaningful trend following the different doses and sampling time points. From a biological point of view, a meaningful trend denotes a compound that responds similarly at all doses and follows a smooth trend along the time points. This type of information can be used to distinguish relevant metabolites from those compounds not following the expected trends. The method is based on analysing the time and dosage trends of each compound via principal component analysis. As only local information is analysed at a time (meaning no correlation with other metabolites is taken into account), the proposed model flags relevant metabolites even if their trend is different from that of any other compound. The new method is therefore an attractive way to reduce the long list of detected compounds in a metabolomics sample set to include only those having the expected smooth time profile that is common for all doses. The new strategy is tested on a sample set obtained from a gut fermentation study of a polyphenol-rich diet. For this study, the initial list of over 25,000 potentially interesting features was reduced to less than 250, thus significantly reducing the expensive and time-consuming manual examination.