Clustering

We clustered the gene expression data matrix using hierarchical clustering, k-means, clusternomics – an integrative context-dependent clustering for biomedical datasets [58], EMMIXgene – a mixture model-based approach to cluster microarray expression data [59], gama – a genetic approach to maximize a clustering criteria [60], DIANA – a divisive, not agglomerative, hierarchical clustering, FANNY – a fuzzy clustering approach, and PAM – partitioning around medoids [61]. The implementations were done in R v 3.6.3 using the packages (functions) stats (hclust, cutree, k-means) v 4.2.0, clusternomics v 0.1.1, EMMIXgene v 0.1.3, gama v 1.0.3, cluster (diana, fanny, pam) v 2.1.2. Validation of the clustering results was performed using the biological homogeneity index (BHI) [62], V-measure [63], and adjusted Rand index (ARI) [64]. These metrics were calculated using R packages clValid v 0.7, saber v 0.3.2, and mclust v 5.4.7, respectively.