Fig. 4

Independent validation of candidate metabolic drivers in the Metabric breast cancer cohort (n = 1991). a The overlap of candidate metabolic drivers identified in the TCGA BRCA cohort and tested for the correlation between mRNA and gene copy number data (log₂ ratio) in the Metabric cohort. The numbers indicate correlated genes in the intrinsic subtypes of breast cancer (PAM50). Due to absence of subtype-specific candidate metabolic genes in the PAM50 subtype ‘Normal-like’ breast cancer, it was not considered in subsequent analyses. b Genome-wide altered copy number fraction in the complete Metabric cohort (n = 1991). Bottom: pink peaks indicate fractional copy number gains/amplifications; blue peaks indicate fractional homozygous or heterozygous deletions. The gene symbols of candidate metabolic genes are located at the top in the order of their genomic location (left to right: chromosome 1 to X). Gene symbols with asterisks indicate significantly altered known breast cancer genes [25], MYC and ERBB2. Genes in red represent clusters of metabolic and significantly altered breast cancer genes that are in genomic proximity of each other. The red dashed lines show their approximate loci. Top: heat map showing the presence (blue) and absence (grey) of candidate metabolic genes across breast cancer intrinsic subtypes (PAM50) (as summarised in Fig. 4a). The subtype-specific significantly altered known breast cancer genes [25] are also highlighted with asterisks. c Correlation between mRNA and copy number data for genes in chromosome 8q24 amplicon using the complete Metabric cohort (All) and intrinsic subtypes of breast cancer. The candidate metabolic drivers are highlighted with unique symbols to show their mRNA dependence on gene dosage. d Copy number-based Kaplan–Meier analysis of SQLE in the Metabric breast cancer cohort. There were only five cases for which copy number state = heterozygous loss and, therefore, these were merged with the copy number diploid (NEUT, n = 1282) group. Genomic gains and amplifications were collapsed into one group (GAIN). e Same as (d) using the MYC diploid/loss subset of the Metabric breast cancer cohort. f mRNA-based Kaplan–Meier analysis of SQLE in the Metabric breast cancer cohort. Samples were split into four groups based on 75th percentile, median and 25th percentile of log₂ mRNA abundance of SQLE (lowest = Q1, highest = Q4). g mRNA-based Kaplan–Meier analysis of SQLE in the MYC diploid/loss subset of the Metabric breast cancer cohort. h Median-dichotomised mRNA-based Kaplan–Meier analysis of SQLE further stratified into hypoxia high and low risk groups (S0H0 = low SQLE and low hypoxia, S0H1 = low SQLE and high hypoxia, S1H0 = high SQLE and low hypoxia, S1H1 = high SQLE and high hypoxia). BC breast cancer, OS overall survival