Development of an integrated radioproteomic signature for improved ejection fraction in end-stage kidney disease postkidney transplantation via cardiac MR T1 maps

Cardiovascular disease is the leading cause of morbidity and mortality in patients with chronic kidney disease, and is associated with upto a 50% mortality rate. Myocardial fibrosis, a feature of uremic cardiomyopathy, is a common occurrence in end-stage kidney disease (ESKD) and is a strong independent predictor of major adverse cardiovascular events as well as being associated with changes in ejection fraction (EF). While plasma proteomics have been explored to understand molecular pathways associated with uremic cardiomyopathy and EF, there have not been comprehensive efforts to combine these with structural phenotypes of myocardial fibrosis via cardiac magnetic resonance images (CMR) T1 maps. Toward addressing this challenge, we present initial results of developing a novel radioproteomic signature for reduced EF in ESKD patients. We leveraged a prospectively accrued cohort of ESKD patients (actively listed to receive a kidney transplant), and for whom plasma samples had been obtained together with CMR T1 maps at baseline and after 9 months. Following pre-processing of CMR T1 maps and plasma proteomics data, machine learning analysis was utilized to identify and integrate top-ranked radiomic and proteomic markers associated with an improvement in EF. A random forests classifier based on the integrated radioproteomic signature yielded a cross-validated area under the receiver operating characteristic curve (AUROC) of 0.73 ± 0.06 via baseline scans and 0.71 ± 0.06 via follow-up scans for distinguishing patients who did or did not exhibit improvement in EF, which was significantly improved over using proteomic or radiomic markers individually. Our novel multimodal feature interrogation approach could enable improved structural and biological characterization of myocardial fibrosis in vivo, based on predicting improvement in ejection fraction in ESKD patients.