Research Paper Volume 16, Issue 7 pp 6314—6333

Machine learning identifies novel coagulation genes as diagnostic and immunological biomarkers in ischemic stroke

class="figure-viewer-img"

Figure 3. Consensus clustering of IS patients based on DECGs. (A) Delta area plot displaying the relative change in area under CDF curve between K and K-1. (B) Cumulative distribution function (CDF) plot showing consensus clustering under K = 2, 3, 4, 5, 6, and when K = 2, the classification is stable. (C) Consensus matrix at K = 2. The values of the consensus matrix are shown in white to dark blue from 0 (impossible to be clustered together) to 1 (always clustered together), and the consensus matrix was arranged according to the consensus clustering (dendrogram above the heatmap). The bars between the dendrogram and heatmap represent the molecular subtypes. All results in A-C indicate that the sample clustering was stable and robust that the boundary of the consensus matrix was clear. (D) Box plot showing the difference of the proportions of 28 immune cells between cluster 1 and cluster 2. In the x-axis, the significant differentially infiltrated immune cells were marked in red color. The value shown in the y-axis was the enrichment score of immune cells calculated by ssGSEA. *p <0.05, nsp > 0.05.