Extended Data Figure 4: Gene-expression patterns and fusion genes found in cervical cancer.

a, Hierarchical clustering (uncentred correlation with centroid linkage as the clustering method) was performed on 4,039 expressed and highly variable genes across samples from 178 cervical, 170 endometrial and 279 head and neck cancer patients. Normalized gene-level RSEM values were median-centred before clustering and relative increased expression values are indicated in red and relative decreased expression values are indicated in blue. Samples from patients with cervical (CESC, light blue), endometrial (UCEC, purple) and head and neck (HNSC, orange) cancer are categorized by different colours as indicated. Also included are indications of HPV status, histology of cervical and endometrial cancers, and tissue site for head and neck cancer samples. Select genes are noted to the right of their locations on the heatmap. b, Box plots of the three differentially expressed SMGs and top six significantly differentially expressed non-SMGs across the iCluster groups using Kruskal–Wallis test. All genes are significantly different between the keratin-low and keratin-high clusters. Significant P values across keratin-low and keratin-high clusters are presented. c, A schematic of BCAR4 tandem duplication in one case (C5-A3HF), detected by analysis of somatic copy number (top) and structural variation (middle). Split reads and genomic breakpoints indicating the tandem duplication are shown. At the RNA level (bottom) the last exon of BCAR4 forms a fusion gene with the first exon of ZC3H7A (red bars indicate the location of mRNA breakpoints; NR_024049 shown as BCAR4 representative transcript). d, Schematic of recurrent fusions (CPSF6–C9orf3, ARL8B–ITPR1 and MYH9–TXN2) or fusions with known occurrences in other cancer types (FGFR3–TACC3), detected by at least two RNA-seq fusion callers in 178 samples. Red bars indicate the mRNA breakpoints.