Identification of patient subtypes based on protein expression for prediction of heart failure after myocardial infarction

Figures

• 

  Figure 1. Overview of the study REVE-1 study5 was used as derivation cohort and REVE-2 study6 as validation cohort. Proteomic data analysis was performed on all plasma samples collected during hospitalization of patients from both cohorts by SOMAscan assay (version V4.0). The SOMAScan platform measured accurately 4668 plasma proteins. Log2 transformed data were centered and reduced. Standardization parameters were calculated with the data from the REVE-1 cohort (see Table S2) and then applied to both REVE-1 and REVE-2 cohorts. Univariate competing risk models were fitted for each protein, and significance tests selected 50 proteins to be associated with occurrence of hospitalization for HF (p< 1 · 07 10−5 in accordance with Bonferroni’s method). Clustering was then performed on the REVE-1 patients with the 50 selected proteins, by k-means procedure with k = 2 groups. Patients of REVE-2 were then assigned to one of the groups identified on REVE-1. A competing risk model was then developed based on the group information on both cohorts.

• 

  Figure 2. Heatmap visualization of the two identified groups on REVE-1 (A) and REVE-2 (B) cohorts Columns represent the 50 selected proteins significantly modulated according to the outcome of patients and classified in the same order for both cohorts. Lines represent the patients divided in the two groups (G1 and G2) identified by clustering with the k-means. Cells are colored according to protein abundance expressed in RFU (Relative Fluorescence Unit). Red represents high abundance whereas blue indicates low abundance. Protein levels are log2-transformed and standardized. See also Table S4.

• 

  Figure 3. Cumulative incidence curves of HF for patients from REVE-1 (A) and REVE-2 (B) cohorts The studied populations are divided in two groups according to the two identified groups by clustering with the k-means. See also Table S3 and Figure S3.

• 

  Figure 4. IPA based protein networks involved in worse outcome in post-MI patients of REVE-1 study The IPA analysis was performed on the 50 proteins selected to be associated with long term survival and listed in Table 2 (version: 73,620,684 Release: 2022-03-12). Only network 1 (A) and network 2 (B) from Table 4 are presented. Nodes are displayed using various shapes that represent the functional class of the proteins as published on https://qiagen.secure.force.com/KnowledgeBase/KnowledgeIPAPage?id=kA41i000000L5rTCAS. The color of proteins indicates respectively their regulated expression (red: increased and green: decreased) associated with occurrence of hospitalization for HF. The arrows indicate the modulatory effect of protein on its interacting proteins. Only direct interactions were selected. Detailed information on the molecules present in the networks is detailed in Table 4. See also Table S7.