The deconvolution and identification were performed using Mass Hunter Quantitative Unknowns Analysis software (B.07.00, Agilent, Santa Clara, CA, USA), alignment with Mass Profiler Professional software (version 13.0, Agilent, Santa Clara, CA, USA), and peak integration using Mass Hunter Quantitative Analysis software (version B.07.00, Agilent, Santa Clara, CA, USA). The identification was performed mainly based on the accurate mass and product ion spectrum matching using the in-house library of authentic standards as well as Fiehn’s and NIST 14 libraries.
In order to perform the differential analysis of the metabolomics data, the variables were then filtered as described by Godzien et al. [46 (link)]. Missing values were replaced by k–means nearest neighbour [47 (link)] using the in-house built scripts for MATLAB 7.10 R2010a (MathWorks Inc., Natick, MA, USA).
Before the statistical analysis, clinical sample areas were normalized by IS abundance to minimize the response variability coming from the instrument. Finally, data were filtered based on the coefficient of signal variation (CV) in QC samples, considering values lower than 30% as acceptable.
In order to perform the differential analysis of the metabolomics data, the variables were then filtered as described by Godzien et al. [46 (link)]. Missing values were replaced by k–means nearest neighbour [47 (link)] using the in-house built scripts for MATLAB 7.10 R2010a (MathWorks Inc., Natick, MA, USA).
Before the statistical analysis, clinical sample areas were normalized by IS abundance to minimize the response variability coming from the instrument. Finally, data were filtered based on the coefficient of signal variation (CV) in QC samples, considering values lower than 30% as acceptable.