Extended Data Fig. 8: Clustering of RNA-seq samples.
From: Deficient H2A.Z deposition is associated with genesis of uterine leiomyoma
a, log2FC of genes for which decreased (FDR < 0.05, FC < −1), increased (FDR < 0.05, FC > 1) or no-change H2A.Z peaks are located at the TSS. log2FC measured by differential expression analysis of tumours (MED12 (n = 38), HMGA2 (n = 44), HMGA1 (n = 62), FH (n = 15), OM (n = 15) and YEATS4 (n = 16)) against normal myometrium (n = 162). H2AZ binding differences are from the spike-in ChIP experiments comparing MED12 (n = 2), YEATS4 (n = 2), HMGA2 (n = 2), HMGA1 (n = 4), OM (n = 2) and FH (n = 2) tumours against normal samples (n = 4). Boxplots show the median and the first and third quartiles. Error bars extend up to 1.5× IQR beyond the quartiles. b, Heatmap presentation of 426 genes that separate myoma subclasses, selected on the basis of linear discriminant analysis. The ordering of samples and genes is based on an unsupervised hierarchical clustering of the 5% (n = 1,355) most variable genes. Two genes per gene cluster are highlighted on the basis of the highest absolute value in discriminant vectors. Patients from whom more than one tumour entered the analysis are highlighted in separate colours. All 426 genes are presented in Supplementary Table 20. c, Consensus clustering of RNA-seq samples. x-axis is sorted by subclass (from left to right: FH, HMGA1, HMGA2, MED12, normal myometrium, OM, YEATS4, and unknown; subclass labels are shown for reference). Item consensus is the mean consensus of an item with all the other items in the same cluster. For each sample, the item consensus value corresponding to each cluster (k = 26) is represented by a colour. For example, all FH samples have the largest item consensus on cluster 2, represented by dark green. Both YEATS4 and OM samples cluster predominantly to cluster 12, represented by blue. Unknown samples form several small clusters. The item consensus value (ei) of each cluster (k) is presented on the y-axis. It is defined as: \({m}_{i}(k)\,=\,\frac{1}{{N}_{k}-1\{{e}_{i}\in {I}_{k}\}}\sum _{j\in {I}_{k},j\ne i}M(i.j)\), where M is distance matrix and Nk is the number of items in the cluster. See Monti et al.55 for further details.
