Predictive metabolomics for detection, interpretation and validation of metabolite patterns in human cerebrospinal fluid
(English)Article in journal (Other academic) Submitted
We here present our predictive metabolomics approach for screening and comparing metabolomics data from human cerebrospinal fluid (CSF) generated by gas chromatography-time of flight mass spectrometry (GC-TOFMS). The approach is based on a combination of hierarchical multivariate curve resolution (HMCR) and manual integration of the GC–TOFMS data for quantification and identification of metabolites in multiple CSF samples. Chemometric data analysis, orthogonal partial least squares (OPLS), for multiple CSF sample comparisons. We show how the predictive feature of both HMCR and OPLS can be used for biomarker detection and verification as well as for diagnostic modelling. To exemplify the capability of the method we have used human CSF from two test subjects aliquoted into 44 tubes stored at either -80 °C or -20 °C as a model system. A total of 170 potential metabolites were resolved from the GC-TOFMS data using HMCR. OPLS modelling revealed a clear separation of the samples according to storage temperature, with a prediction accuracy of 100% using a test set.
cerebrospinal fluid, metabolomics, predictive metabolomics, GC-MS, chemometrics
IdentifiersURN: urn:nbn:se:umu:diva-26871OAI: oai:DiVA.org:umu-26871DiVA: diva2:274604