Progenesis qi

LC-MS raw data were deconvolved using Progenesis QI (Waters Corp., Milford, Massachusetts, United States). Peak picking, alignment, and area normalisation were carried out using one of the QC data files as the reference. Significant features extracted from raw data were aligned to significant features in the reference sample, using a RT window ± 0.2 min and mass tolerance ± 5 ppm filters. Features were annotated using accurate mass match and tandem MS data with the human metabolome database (HMDB). Mass tolerances of 5 and 5 ppm were applied for precursor and fragment ions, respectively. Compounds with a fragmentation score <20 were not annotated. Progenesis QI score, fragmentation score, and isotope similarity were reported for all annotations based on a combination of accurate mass and fragmentation data, seen in Supplementary Table S1 concluding 3461 metabolite features in negative ion mode and 2426 metabolite features in positive ion mode.

Baseline filtering, peak identification, integration, retention time correction, and peak alignment were completed through Progenesis QI (Waters Corporation, Milford, USA). Finally, the data matrix of retention time, mass–charge ratio and peak intensity was obtained. Data analysis was performed on the Majorbio Cloud Platform (https://cloud.majorbio.com) to upload data for subsequent analysis, data pretreatment, removing QC samples Relative standard deviation (Relative standard deviation, RSD) > 30% of the variables, and log10 logization process, to obtain the final data matrix for subsequent analysis.

Data alignment and peak detection were performed in Progenesis QI (Nonlinear Dynamics; Waters Corp.) with normalization to all compounds. Retention time calibration and lipid identification were calculated with the Python package LiPydomics (66 (link)). Multivariate statistics were created through LiPydomics and ClustVis (66 (link), 67 (link)). MS/MS analysis and identification of the most abundant FAs were performed in Skyline utilizing a targeted lipid library generated with LipidCreator (68 (link), 69 (link)).

The raw data were imported into the Progenesis QI software (Waters Corporation, MA, USA) for alignment. Further, the peak picking and identification of polar lipids were carried out using a high-resolution positive-ion MS, and the absolute intensities of all identified compounds were recalculated to determine the relative abundances and to normalize the values of the lipid molecules. The data were then exported into the EZinfo 2.0 software (Sartorius Stedim Biotech, Umeå, Sweden) for the multivariate statistical analysis, and the principal component analysis (PCA) and orthogonal projections to latent structures discriminant analysis (OPLS-DA) were used to create the final statistical models to obtain the group clusters. Lipid molecules with the strongest effect on the group clustering were identified as those with a VIP greater than one. In addition, potential ions with p < 0.05 and FC > 2 were selected for further metabolite relationship pathway characterization using the MetaboAnalyst web server. Human Metabolome Database (HMDB) IDs were matched with the Kyoto Encyclopedia of Genes and Genomes (KEGG) IDs for the KEGG mapping. IDs without a match were excluded from the analysis, and the Rattus norvegicus (rat) pathway library in the KEGG was selected for analysis.

The acquired mass data were imported to Progenesis QI (Waters Corporation, Milford, MA, USA) for peak detection, alignment, deconvolution, peak picking, and normalization. Then a three-dimensional data matrix was output composed of the sample name, peak number (t_R-m/z pair), and ion intensity. Finally, the resulting matrix was imported into SIMCA (version 14.1, Umetrics AB, Umeå, Sweden) for multivariate statistical analysis such as principal component analysis (PCA) and orthogonal partial least squares discriminant analysis (OPLS-DA) to classify the metabolic phenotypes. The data quality control was completed with SIMCA (Details are listed in Additional file 1). The loading plot from OPLS-DA together with the variable importance in the projection (VIP) was used to discover the potential differential compounds. Hierarchical cluster analysis (HCA) was conducted to estimate the consistency of these drugs. The HCA heatmap analysis shows the change in the content of all ions in each sample by a gradient of color change (blue-white-red).

Raw data acquired from GC/MS and UPLC-IMS-QTOF/MS were pre-processed by LECO Chroma TOF and Waters Progenesis QI, respectively. The raw data was performed in the pre-process procedures, including peak picking, baselining and alignment. Then, the pre-processing data matrix containing feature name (named as retention time and m/z,), sample information (four biological replicates per sample), relative abundance (calculated by peak area) was prepared and submitted to MetaboAnalyst (https://www.metaboanalyst.ca/). After that, the data matrix was performed three categories of normalization, including normalization by median, log transformation and auto-scaling via online data analysis software MetaboAnalyst embedded algorithm.