Skip to main content
Fig. 1 | BMC Biology

Fig. 1

From: Prevalence, causes and impact of TP53-loss phenocopying events in human tumors

Fig. 1

Evaluation of the TP53 function loss score classifier and prevalence of TP53 loss phenocopying events in cancer. a Receiver operating characteristic (ROC) curve and area under the ROC (AUROC) curve for training and testing sets in TCGA tumor transcriptomes. b Bottom: FDR for each tumor sample. X axis shows the classification score thresholds for each tumor sample. The general threshold used for classification (0.6) is highlighted. Top: the histogram of frequency of CNV events (“density” refers to smoothed relative frequency) affecting TP53 and the known phenocopying genes MDM4, MDM2 and PPM1D at various phenocopy score thresholds. c TP53 loss phenocopying score stratified by 3 known phenocopying CNA events and by TP53 deletions. Data points are tumor samples coloured by TP53 status; boxes show median, Q1 and Q3, while whiskers show range (outlying examples shown as separate dots). X axis represents the GISTIC thresholded CNV of each given gene. Tumor samples with deletions in the corresponding genes (for MDM2, MDM4 and PPM1D) and amplifications (TP53) are omitted for simplicity. P values are by t-test comparisons of the TP53 phenocopy score between each shown CNV category to the neutral CNV (0) category in TP53 wild-type samples. d TP53 functional status distribution across TCGA cancer types. Left: pan-cancer; “Phenocopy” refers to TP53-loss phenocopying tumors according to the classifier in panels a and b. Right: showing only the TP53 loss phenocopying tumor samples by cancer type, further stratified by cause of the phenocopy. Tumor samples harbouring a known event that affects TP53 functionality are shown with colours, and the remaining TP53-loss phenocopy tumors are labelled as “Unknown cause”. KIRC and DLBC were omitted as they only had 1 phenocopied tumor sample (”unknown cause”)

Back to article page