Whole blood transcriptome profile at hospital admission discriminates between patients with ST-segment elevation and non-ST-segment elevation acute myocardial infarction

Figures

• 

  Figure 1. Flow diagram of the study.

• 

  Figure 2. Differential gene expression in STEMI vs. NSTEMI matched patients. Statistical analysis was performed by negative binomial Generalized Linear Model, controlling for multiple testing using the false discovery rate (FDR) by the Benjamini-Hochberg procedure. (a) Heatmap depicting relative expression abundance of differentially expressed (DE) genes (FDR-adjusted P < 0.05) in STEMI (n = 15) vs. NSTEMI (n = 15) patients, matched for age, sex, and cardiovascular risk factors. Unsupervised average-linkage hierarchical clustering based on Spearman dissimilarity matrix allowed complete separation between STEMI and NSTEMI, suggesting that these 323 DE genes strongly associated with the specific AMI phenotype. Gene expression levels were expressed as log2 transformed normalized counts and displayed as gradient colours from higher (dark orange) to lower (dark blue). (b) Volcano plot depicting log2 mean fold-differences (STEMI vs. NSTEMI, n = 15 for both groups, x-axis) versus −log10 P-values (y-axis) of all genes, stemming from the differential analysis not corrected for the level of cardiac troponin I (cTnI) on admission. Significant DE genes are coloured: 151 annotated (red dots) and 29…

• 

  Figure 3. Enrichment map of gene-sets stemming from the analysis on the cTnI-corrected dataset. Functional enrichment investigation on genome-wide expression profiles was done by Gene Set Enrichment Analysis (GSEA), using as gene ranking metric the likelihood ratio statistics of the differential expression analysis performed by GLM, correcting for cTnI on admission, in STEMI (n = 15) vs. NSTEMI (n = 15) patients matched for age, sex, and cardiovascular risk factors. To visually interpreting GSEA results, a network of the most significant Gene Ontology biological processes (GO-BP; at an FDR-adjusted P < 0.05) was drawn. The node colour associates with STEMI (red) or NSTEMI (blue) phenotype; node gradient colour is proportional to node significance, from lower (light) to higher (dark); node size is proportional to the gene-set size. Edge thickness is proportional to the similarity between two gene-sets, for a cut-off of 0.25 of the combined Jaccard plus Overlap coefficient. An extended list of GO-BP gene-sets significant at a nominal P-value < 0.05, along with enrichment statistics, is given in Supplementary Table S2a.

• 

  Figure 4. Enrichment map of gene-sets unique to the admission cTnI-corrected dataset. Functional enrichment analyses were done by GSEA, using as gene ranking metrics the likelihood ratio statistics of the differential expression analyses performed by GLM either correcting or not for cTnI on admission (in STEMI vs. NSTEMI patients, n = 15 for both groups, matched for age, sex, and cardiovascular risk factors). The map derived by subtracting the GSEA results obtained using the uncorrected gene ranks from those on the cTnI-corrected gene ranks. The colour scheme is as in Fig. 3. The complete lists of GO-BP gene-sets significant at a nominal P-value < 0.05, along with enrichment statistics, in either the uncorrected or the cTnI-corrected models, are given in Supplementary Table S2a.

• 

  Figure 5. Genes predicting the cTnI peak. Linear regression was used to model the relationship between peak cTnI level (as the outcome variable) and peripheral blood gene expression on hospital admission (explanatory variable). Scatter plots and trendlines show six of the most significant genes that positively or negatively correlate with cTnI peak (see Table 2 for statistics). The full list of genes significantly predicting cTnI peak at an FDR-adjusted P-value < 0.05, along with the coefficients of determination R2, the regression β coefficients, and the exact P-values, in either the uncorrected or the cTnI-corrected models, is given in Supplementary Table S3. STEMI (n = 15) and NSTEMI (n = 15) samples are highlighted by red and blue dots, respectively. The 95% confidence interval of the trendline is depicted in light green.

• 

  Figure 6. Validation of top-ranked genes on an independent cohort of consecutive patients. The expression level of 8 significant genes, chosen among the top DE genes and/or the top cTnI peak-associated genes, was evaluated in patients from the validation cohort using RT-qPCR single assays. Pearson’s correlation coefficient (r) was computed to assess the degree of association between the log2 mean fold-differences (log2FC) of these genes observed in the discovery group of matched patients (STEMI n = 15 vs. NSTEMI n = 15) and the log2FC found in the validation cohort (STEMI n = 45 vs. NSTEMI n = 43). Data are plotted as log2FC STEMI vs. NSTEMI in the study (y-axis) versus the validation patient cohort (x-axis). The 95% confidence interval of the trendline is depicted in light green.