Fig. 5: Inactivation mechanisms in HRD tumors and associations with patient outcome.
From: Homologous recombination deficiency in primary ER-positive and HER2-negative breast cancer
a All 42 HRD tumors. b Six HRD PAM50 Basal tumors. c Six HRD PAM50 HER2E tumors. d 27 HRD PAM50 LumB tumors. In a–d, a few cases exist harboring both promoter methylation and a somatic variant for a gene. For these cases, only one of the alterations has been counted as detailed in Supplementary Data 5, based, e.g., on correlation to mRNA expression. e Principal component analysis in all HRD tumors (left) and in LumB HRD tumors (right) using the 5000 most variant genes based on FPKM data for the respective group, colored by the proposed HRD inactivation mechanism. Principal components 1 and 2 (PC1 and PC2, respectively) are shown. The LumA, Basal, and HER2E groups are excluded due to a few cases. f Kaplan-Meier plot of the association of HRD status with DRFI for patients with HRD and HR-proficient WGS analyzed tumors treated with Endo. DRFI was based on cancer registry data from Staaf et al.27. Univariate Cox regression hazard ratio and 95% confidence interval shown. g Kaplan-Meier plot of the association of HRD status with DRFI for HRD and HR-proficient WGS analyzed tumors with inclusion of 2046 additional non-overlapping Endo-treated patients with unknown HRD status from Staaf et al.27 stratified by their PAM50 LumA status (LumA or not-LumA). DRFI was based on cancer registry data from ref. 27 for non-WGS patients. h Kaplan-Meier plot of the association of HRD status with DRFI for patients with HRD and HR-proficient WGS analyzed tumors treated with ChemoEndo. Clinical review: DRFI data were used as the endpoint. Univariate Cox regression hazard ratio and 95% confidence interval shown. i Kaplan-Meier plot of the association of HRD status with DRFI for HRD and HR-proficient WGS analyzed tumors with inclusion of 761 non-overlapping ChemoEndo treated patients with unknown HRD status from Staaf et al. 27 stratified by their PAM50 LumA status (LumA or not-LumA). For the 761 patients, the DRFI data were based on cancer registry data from ref. 27, while for the WGS analyzed samples, clinical review data were used as the endpoint. In Kaplan-Meier plots, the p-value was calculated using the log-rank test.
