To systematically investigate the metabolic characteristics of hemorrhagic shock (HS) under simulated high-altitude conditions by integrating metabolomics and machine learning algorithms, and screen the biomarkers for predicting HS under these conditions.

Eighty SPF-grade male SD rats (10 to 12 weeks old, weighing 180 to 220 g) were randomly divided into a sham operation group (Sham) and an uncontrolled HS group (UHS), with 40 animals in each group. All rats were placed in a hypobaric hypoxia chamber to simulate an altitude of 5 000 m for 48 h. After removal from the chamber, the rats were given an injection of oleic acid immediately via tail vein, and then in 30 min latter, those from the UHS group were induced to free bleeding by splenic artery transection, with the endpoint set at a mean arterial pressure (MAP) of 40 mmHg. Thus, a rat model of HS under stimulated high-altitude conditions was established. Metabolomic analysis was performed on the serum samples to identify differential metabolites between the Sham and UHS groups. Weighted gene co-expression network analysis (WGCNA) was carried out to identify metabolites associated with UHS. Three machine learning methods, including least absolute shrinkage and selection operator (Lasso) regression, random forest (RF), and support vector machine-recursive feature elimination (SVM-RFE) were employed to identify relevant biomarkers for HS at high-altitude. Based on 10-fold cross-validation, the diagnostic performance of the biomarkers was evaluated with receiver operating characteristic (ROC) curve analysis, and the area under the curve (AUC) was calculated.

Principal component analysis (PCA) of the metabolomic data showed clear separation between the samples from the Sham and UHS groups. Orthogonal partial least squares-discriminant analysis (OPLS-DA) further confirmed significant differences in metabolic profiles between the 2 groups. There were 5 398 metabolites identified, with the UHS group having 391 metabolites significantly down-regulated (VIP>1, FC<1/1.5, P<0.05) and 1 181 metabolites obviously up-regulated (VIP>1, FC>1.5, P<0.05) when compared to the Sham group. Among the metabolites, the most altered metabolites were lipids, including FFA (18:2), FFA (18:1), FFA (12:0), FFA (14:0), ergosterol, mesaconic acid, suberic acid, gamma-linolenic acid, and PC [22:5 (4Z, 7Z, 10Z, 13Z, 16Z)/22:6 (4Z, 7Z, 10Z, 13Z, 16Z, 19Z)], and then followed by organic acids and derivatives, such as methionine, citrulline, creatine, L-lactic acid, oxalacetic acid, and 2-acetolactate. Pathway enrichment analysis showed that these differential metabolites were mainly involved in glutamine metabolism, taurine metabolism, alanine, aspartate and glutamate metabolism, as well as glycerophospholipid metabolism. WGCNA analysis found that MEturquoise module exhibited the strongest positive correlation with UHS (R=0.95, P=8e-40), while the MEsalmon module showed the strongest negative correlation with UHS (R=-0.67, P=2e-11). With thresholds of |Gene Significance (GS)|>0.2 and |Module Membership (MM)|>0.8, 344 characteristic metabolites were identified, and all of them were significantly associated with HS in a high-altitude environment. The 3 machine learning algorithms yielded 3 biomarkers, that is, methionine, D-inositol-4-phosphate and phosalone. Significantly increased D-inositol-4-phosphate and phosalone while decreased methionine were observed in the UHS group than the Sham group. ROC curve analysis revealed that the AUC values of D-inositol-4-phosphate, phosalone and methionine were 0.988, 0.950 and 0.988, respectively, indicating that the 3 biomarkers having good predictive efficiency for HS under simulated high-altitude condition.

HS rats under simulated high-altitude conditions present significantly disturbed metabolic profiles, characterized by substantial changes across multiple pathways. The metabolites D-inositol-4-phosphate, phosalone and methionine may serve as potential biomarkers for predicting the occurrence or evaluating the severity of HS in high-altitude environments.