Quanlynx 4

UPLC-MS/MS analysis was carried out on an Acquity-Xevo TQ system (Waters, Barcelona, Spain). The conditions used were: ionization in negative mode (ESI-), capillary voltage 3.5 kV, source temperature 120 °C, desolvation temperature 300 °C, gas flow of the nitrogen cone of 150 L/h, and desolvation flow of 680 L/h.
Separation conditions were selected to achieve appropriate chromatographic retention and resolution by using a C18 column (2.1 × 50 mm, 1.7 μm) (Acquity UPLC BEH) and pre-column (2.1 × 5 mm) from Waters. A binary mobile phase CH₃OH (0.1% v/v HCOOH):H₂O (0.1% v/v HCOOH) with gradient elution was used. The flow rate was 0.4 mL/min, the temperatures of column and the autosampler were 37 °C and 4 °C, respectively. The injection volume was 10 µL. The gradient started with 30% v/v CH₃OH (0.1% v/v HCOOH) (i.e., channel B) and from 1 to 4.0 min %B increased up to 90%. Finally, the mobile phase composition returned to the initial conditions at 4.1, and it was maintained for 3.9 min for system conditioning.
The detection was performed by multiple reaction monitoring using the acquisition parameters obtained in a previous study [18 (link),19 (link)].
For data acquisition and processing, MassLynx 4.1 and QuanLynx 4.1 softwares from Waters (Waters, Barcelona, Spain) were used, respectively. Linear response curves were calculated employing PGF_2α-d₄ as internal standard.

Data processing and quantification were performed using the QuanLynx 4.1 software (Waters Corporation, Milford, MA, USA) provided with the instrument. Calibration was performed through linear regression with reciprocal fit weighting to ensure maximum accuracy at the lower concentration range.

Raw data were acquired and processed using MassLynx 4.1 and QuanLynx 4.1 (Waters, Milford, MA, USA), respectively. Linear response curves were obtained from UPLC-MS/MS peak area measurements employing betaine-D₁₁ as internal standard. Further data processing was carried out in Matlab 2015a from Mathworks Inc. (Natick, MA, USA) using the PLS Toolbox 8.0 from Eigenvector Research Inc. (Wenatchee, WA, USA) and in-house written functions. ROCs and AUCs were computed employing MetaboAnalyst 3.034 (link). Missing values were estimated using k-nearest neighbors and data were autoscaled. For multivariate ROC curve based exploratory analysis all available features at each time point were employed. Feature ranking was based on univariate AUC values and random forests were used as a classification method. ROC curves were generated by Monte-Carlo cross validation (MCCV) using balanced subsampling where in each MCCV two thirds of the samples were used to evaluate the feature importance. Then, the model was validated using one third of the samples that were left out during model generation. For the calculation of the CI, this procedure was repeated 500 times.