Skip to main content
Fig. 9 | Genome Biology

Fig. 9

From: Importance of rare gene copy number alterations for personalized tumor characterization and survival analysis

Fig. 9

Impact of rare and frequent patient-specific gene CNAs on survival. Kaplan–Meier curves for patients with negative (blue) and positive (red) average impact of their corresponding tumor-specific gene CNAs on cancer type-specific survival signature genes. Shown are results for independent tumor cohorts (Rembrandt: GBM; CLCGP: LUAD; TCGA: SKCM new patients) that were not involved in any step of CCTN network inference or RF-based prediction of survival signature genes. CCTN derived from cancer cell lines and cancer type-specific survival signature genes identified from TCGA cohorts were used to investigate the impact of rare and frequent gene CNAs on patient survival for these cohorts. Separation of long- and short-lived patients by CCTN is expected to be predictive if patients with positive average survival impact scores (red) tend to survive longer than patients with negative impact scores (blue). The corresponding p value quantifies if the red curve is significantly above the blue curve in comparison to random class label permutations. a–d GBM results. a Patients with positive average impact scores on survival signature genes (red) tend to survive significantly longer than patients with negative impact scores (blue). b Frequent patient-specific gene CNAs (frequency ≥1 % in corresponding training cohort) alone cannot explain the significant difference in (a). c Rare patient-specific gene CNAs (frequency <1 %) significantly contribute to the observed significant differences in (a). d Loss of patient separation into short and long survival for patient-specific survival impact scores computed based on all patient-specific gene CNAs under a random network. e–h LUAD results. Similar to GBM, but in contrast to GBM, frequent mutations alone contribute to a significant separation between long- and short-lived patients. i–l SKCM results. Overall trends are comparable with those observed for GBM. CLCGP Clinical Lung Cancer Genome Project, CCTN cancer cell transcriptional regulatory network, CNA copy number alteration, GBM glioblastoma multiforme, LUAD lung adenocarcinoma, RF random forest, SKCM skin cutaneous melanoma, TCGA The Cancer Genome Atlas

Back to article page