Integration of transcriptomics and metabolomics identify biomarkers of aberrant lipid metabolism in ulcerative colitis

Abstract

Background: The incidence of ulcerative colitis (UC) continues to rise globally, but effective therapeutic targets are still lacking. In recent years, numerous studies have indicated that lipid therapies could offer a novel perspective for UC treatment. Given the absence of prior research utilizing high-throughput data to identify target genes associated with lipid metabolism, we conducted this work.

Methods: The training set for this study was derived from four datasets within the Gene Expression Omnibus (GEO), encompassing a total of 357 UC patients. We employed four machine learning methods (LASSO, SVM, RF, and Boruta) to jointly identify core biomarkers in these patients, whose aberrant expression needed to be validated in independent datasets and in dextrose sulfate sodium salt (DSS)-induced UC mouse models. Regarding metabolomics, we detected abnormal oxidized lipids in the serum of UC mouse using liquid chromatography-tandem mass spectrometry (LC-MS/MS) in conjunction with orthogonal partial least squares-discriminant analysis (OPLS-DA).

Results: Phospholipase A2 Group IIA (PLA2G2A) was first identified as a possible biomarker for UC, with AUC values of 0.810 and 1.000 in the two validation sets, while in animal models the gene showed similarly significant up-regulation in damaged intestinal mucosa. Further analysis of this gene showed that it was positively correlated with 17 immune cell types and histological severity. Additionally, we pioneered the development of a lipid metabolism score in UC research, which outperformed all individual genes in terms of disease diagnostic efficacy (AUC values of 0.980 and 1.000 for the two validation sets, respectively). Finally, the metabolomics study also identified 31 significantly abnormal oxidized lipids, including 12-HHT and DHA.

Conclusions: PLA2G2A is a key therapeutic target for UC, and oxidized lipids such as 12-HHT can serve as potential serologic indicators for diagnosis.

Keywords: Lipid metabolism; Machine learning; Metabolomics; Oxidized lipids; PLA2G2A; Ulcerative colitis.