Data processing and statistical analysis
After the data acquisition using LTQ-Orbitrap, the raw data were imported and processed (peak alignment, detection and identification) using the Compound Discoverer 2.1 software system. The retention time and m/z data for each peak were determined by the software. For plasma metabolomics and lipidomics analyses, all data were normalized using the peak areas of internal standards. For metabolomics and lipidomics profiling of tumor tissue samples, all data were normalized to the total protein concentrations. All multivariate data analyses were performed using SIMCA version 14.1 software (Umetrics, Inc., Ume, Sweden) system and MetaboAnalyst 4.0 (https://www.metaboanalyst.ca/). Multivariate data analyses and partial least squares-discriminant analysis (PLS-DA) were used to evaluate the differential metabolites between groups in the plasma and tumor samples, and performed with Pareto scaling. The components with a variable importance in the projection (VIP) value exceeding 1.0 were selected as potential compounds that contributed remarkably to the clustering and showed differences between the groups. Student’s t -test was used to evaluate the statistical significance of each group of metabolic changes and considered significant when the value was less than 0.05.
Putative identification and searching was carried out based on the mass adducts ([M+H]+, [M+Na]+, [M-H]-, etc), mass fragment (MS2) ions and retention time. The accuracy tolerance window of the mass was set to 10 ppm while searching the metabolites. The metabolites were initially searched and determined from online databases such as METLIN (https://metlin.scripps.edu/), HMDB (http://www.hmdb.ca/), and KEGG (http://www.genome.jp/kegg) utilizing the detected m/z value and mass fragmentation patterns. Later, the identified metabolites were further crosschecked with the commercially available standards for the confirmation.