Prognosis and Personalized In Silico Prediction of Treatment Efficacy in Cardiovascular and Chronic Kidney Disease: A Proof-of-Concept Study

Abstract

(1) Background: Kidney and cardiovascular diseases are responsible for a large fraction of population morbidity and mortality. Early, targeted, personalized intervention represents the ideal approach to cope with this challenge. Proteomic/peptidomic changes are largely responsible for the onset and progression of these diseases and should hold information about the optimal means of treatment and prevention. (2) Methods: We investigated the prediction of renal or cardiovascular events using previously defined urinary peptidomic classifiers CKD273, HF2, and CAD160 in a cohort of 5585 subjects, in a retrospective study. (3) Results: We have demonstrated a highly significant prediction of events, with an HR of 2.59, 1.71, and 4.12 for HF, CAD, and CKD, respectively. We applied in silico treatment, implementing on each patient's urinary profile changes to the classifiers corresponding to exactly defined peptide abundance changes, following commonly used interventions (MRA, SGLT2i, DPP4i, ARB, GLP1RA, olive oil, and exercise), as defined in previous studies. Applying the proteomic classifiers after the in silico treatment indicated the individual benefits of specific interventions on a personalized level. (4) Conclusions: The in silico evaluation may provide information on the future impact of specific drugs and interventions on endpoints, opening the door to a precision-based medicine approach. An investigation into the extent of the benefit of this approach in a prospective clinical trial is warranted.

Keywords: cardiovascular events; chronic kidney disease; coronary artery disease; heart failure; personalized medicine; urinary biomarkers.

Figure 1

Urinary peptidomics classifiers and primary outcomes. Kaplan–Meier curves for the primary outcome; classifier scores from the lowest (1) to highest (5) quintile, for risk of heart failure events, as assessed by HF2 (A), coronary artery disease events, as assessed by CAD160 (B), and chronic kidney disease progression, as assessed by CKD273 (C).

Figure 2

HF2 classifier treatment responses. HF2 scores were z-scaled across samples for visualization. Heatmap HF2 classifier treatment responses of 5585 patients (A). The top of the heatmap shows the event information. Samples (columns) were ordered based on the HF2 score prior to in silico treatment, from lower scores (left) to higher scores (right). The zoomed-in heatmap shows the treatment response of HF2 in 10 patients (B). Patients who were already receiving one of the treatments at the beginning of the study are depicted in gray.

Figure 3

CAD-160-marker classifier treatment responses. CAD-160-marker scores were z-scaled across samples for visualization. Heatmap CAD-160-marker classifier treatment responses of 5585 patients (A). The top of the heatmap shows the event information. Samples (columns) were ordered based on the CAD-160-marker score prior to in silico treatment, from lower scores (left) to higher scores (right). The zoomed-in heatmap shows the treatment response of the CAD-160-marker classifier in 10 patients (B). Patients who were already receiving one of the treatments at the beginning of the study are depicted in gray.

Figure 4

CKD273 classifier treatment responses. CKD273 scores were z-scaled across samples for visualization. Heatmap CKD273 classifier treatment responses of 5585 patients (A). The top of the heatmap shows the baseline eGFR value information. Samples (columns) were ordered based on the CKD273 score prior to in silico treatment, from lower scores (left) to higher scores (right). The zoomed heatmap shows the treatment response of the CKD273 classifier in 10 patients (B). Patients who were already receiving one of the treatments at the beginning of the study are depicted in gray.

Figure 5

Schematic depiction of the study design. The relative abundance of 5071 sequenced urinary peptides was investigated using CE-MS. Data on some selected peptides (ID) for 1 subject are shown. Several of these peptides were previously identified as being associated with the respective pathophysiology and combined into classifiers CKD273, CAD160, and HF2. Some of these peptides are labelled with their respective color. In the first step, the patient received a score for progression to event using the predefined urinary classifiers. Of the peptides shown, 3, labelled in bold, were found to be affected by the SGLT2i treatment. In the second step, the abundance of these 3 peptides was adjusted, based on the observed fold change, as a result of the treatment (“in silico treatment”). The classifier score was then re-calculated (labelled *), and the result was compared to the initial scoring, where a decrease in the scoring indicated the benefit of the treatment. In this example, the relevant impact of the SGLT2i treatment on a CKD and HF event is predicted, but not the impact on CAD.