Omics-driven strategy develops a tumor biomarker-derived signature to forecast the occurrence and progression of heart failure.

Tumor biomarkers have been implicated in heart failure (HF). This study aimed to explore the role of tumor biomarkers in the occurrence and progression of HF. An integrative machine learning framework was leveraged to develop a predictive tumor biomarker-derived signature (TBS) based on 11 independent HF cohorts. Pathway enrichment analysis was performed to understand the latent mechanisms of TBS. Mendelian randomization (MR) and molecular docking were used to investigate the therapeutic significance of TBS. TBS, a combination of random forest and gradient boosting model fitting on seven tumor biomarkers (FLT1, CSF1, SPP1, ENO2, GOLM1, GPC3, S100B), demonstrated an excellent performance with higher accuracy and sensitivity. ECM receptor interaction and PI3K-AKT signaling present cross-talk pathways underlying TBS genes. A significant causal relationship between genetic predisposition toward S100B and HF was revealed. Quercetin was predicted to interact with S100B, representing a promising therapy from S100B repression. Using a proteomics dataset as external validation, S100B was found to significantly elevate in HF with an AUC value achieving 0.844. TBS has potential implications for the prediction, risk guidance, and treatment of HF patients. S100B is a pivotal biomarker and therapeutic target for HF.