Integrated proteomics and machine learning for identifying candidate serum biomarkers in acute myocardial infarction-complicated cardiogenic shock: a prospective exploratory study.

Abstract

Background: Cardiogenic shock secondary to acute myocardial infarction (AMI-CS) prohibitively impacts survival. This prospective study aimed to discover and internally verify candidate serum protein biomarkers and evaluate their potential prognostic value for 30-day mortality in AMI-CS patients.

Methods: AMI-CS patients were consecutively enrolled into discovery (n = 30) and verification (n = 60) cohorts. Candidate biomarkers were screened using Data-Independent Acquisition (DIA) mass spectrometry, analyzed via differential abundance and weighted gene co-expression network analysis (WGCNA), and verified via targeted Parallel Reaction Monitoring (PRM). Boruta feature selection for five machine learning algorithms were embedded within a rigorous nested cross-validation scheme. Incremental prognostic value over clinical predictors was evaluated using Cox regression and metrics including the integrated discrimination improvement (IDI).

Results: DIA proteomics identified 216 proteins differentially abundant between 30-day survivors and non-survivors, and WGCNA defined an outcome-associated module linked to shock severity and enriched for oxidative stress and energy metabolism. During PRM verification, leakage-free nested cross-validation random forest model selected a seven-protein panel (YWHAZ, QDPR, MDH2, FAH, PSMA1, FABP5, and AHCY), which achieved a mean area under the ROC curve of 0.82 (95% CI: 0.69-0.95) for 30-day mortality. Incorporating this panel significantly enhanced the IDI of the SCAI model (IDI 0.143, P = 0.016) and the IABP-SHOCK II model (0.138, P = 0.034), which remained independent after adjusting for core clinical confounders.

Conclusion: This exploratory study identifies a seven-protein candidate panel with potential value for early short-term risk stratification in AMI-CS patients. Large-scale multi-center external validation is strictly warranted to confirm its real-world clinical utility.