Plasma proteomics integration reveals foxO and relaxin signaling as primary drivers of myocardial radiomic heterogeneity in end-stage HFrEF
P Laba, M K Marta Kacperska, M K Maciej Mazuruk, P L Przemyslaw Leszek, I M Ilona Michalowska, M M Michal MaczewskiAbstract
Background
Non-invasive assessment of adverse myocardial remodeling in end-stage heart failure with reduced ejection fraction (HFrEF) remains a critical clinical challenge. While computed tomography radiomics provides detailed macroscopic textural phenotyping, its underlying molecular drivers are poorly understood. We aimed to bridge this gap by mapping systemic proteomic signatures to specific myocardial radiomic features in patients evaluated for left ventricular assist device (LVAD) implantation.
Methods
We analyzed a prospective cohort of 43 patients with advanced HFrEF prior to LVAD implantation. Coronary Computed Tomography Angiography (CCTA) and concurrent systemic blood samples were obtained. We extracted 93 radiomic features from the myocardium, encompassing first-order statistics, Gray-Level Co-occurrence Matrix (GLCM), Gray-Level Size Zone Matrix (GLSZM), Gray-Level Run Length Matrix (GLRLM), Gray-Level Dependence Matrix (GLDM), and Neighborhood Gray-Tone Difference Matrix (NGTDM). Plasma proteomics was assessed using the Olink Explore 5400+ platform. Proteomic profiles were mapped to 282 KEGG pathways. To predict radiomic phenotypes from pathway-specific protein expression, we utilized a high-throughput Partial Least Squares (PLS) regression pipeline with 5-fold cross-validation. Model robustness was strictly evaluated using the cross-validated predictive variance metric (Q2), applying a rigorous cutoff of Q2 ≥ 0.30 to minimize false discoveries.
Results
Our multi-omics pipeline successfully identified significant proteomic-radiomic associations. The most robust biological signal originated from the FoxO signaling pathway, which strongly predicted myocardial structural heterogeneity. Specifically, FoxO pathway proteins accounted for 48.1% of the cross-validated variance in first-order energy (Q2 = 0.481). Furthermore, the FoxO pathway significantly predicted total energy (Q2 = 0.413), robust mean absolute deviation (Q2 = 0.345), and interquartile range (Q2 = 0.342). Additionally, the Relaxin signaling pathway demonstrated a strong predictive association with first-order energy (Q2 = 0.320).
Conclusion
Systemic proteomic expression, predominantly within the FoxO and Relaxin signaling pathways, directly correlates with macroscopic myocardial textural heterogeneity in end-stage HFrEF. This robust radioproteomic link suggests that specific CT-derived radiomic features could serve as non-invasive, "virtual biopsy" surrogates for the underlying metabolic and structural remodeling driven by these pathway.sPicture