Fig. 2

Unsupervised clustering using an agglomerative hierarchical clustering method. A Dendrogram of the hierarchical clustering based on differentially expressed genes (DEGs). B Silhouette scores and the sum of squared error were used to identify the optimal number of clusters. The maximum peak of the silhouette score plots and the point at which the sum of squared error begins to diminish (the “elbow” method) determined the optimal number of clusters. C t-distributed stochastic neighbor embedding (t-SNE) and principal component analysis of the DEG profiles. C1 (n = 252) and C2 (n = 333) are colored red and yellow, respectively