Abstract 50: Differences in gene-expression profiles in breast cancer between African and European-ancestry women

Abstract

Abstract Background: African American (AA) women suffer from higher incidence of some forms of breast cancer and higher overall mortality than European American (EA) women. Methods: To investigate the biological basis for the racial difference in breast cancer, we compared gene expression profiles in breast cancer tissues using RNA sequencing data derived from 260 AA and 155 EA women who are participants of the Southern Community Cohort Study (SCCS) and were diagnosed with breast cancer during the cohort follow-up. Genes differentially expressed between AA and EA breast cancer patients were identified using linear regression models thresholded at false discovery rate (FDR) < 0.01. Covariates adjusted in the analysis include age at diagnosis, PAM50 subtypes, TNM stages, probabilistic estimation of expression residuals (PEER) factors, and test batch. Penalized logistic regression via ridge regularization was used to avoid overfitting and overcome the potential problem of multicollinearity by shrinking the model coefficients in building the race-specific gene expression signature. Gene expression and clinical data of 180 AA and 838 EA breast cancer patients from the Cancer Genome Atlas (TCGA) were used for external validation of the gene signature. Penalized Cox regression via elastic net was used to identify a subset of the race-differentiated genes that are associated with breast cancer survival. Results: We identified 59 genes (54 protein-coding genes and 5 long intergenic non-coding RNAs) that were differentially expressed between EA and AA with an FDR adjusted p-value < 0.01. The top three identified genes were IL20A (OMIM 605620), MARCO (OMIM 604870), and BTN3A1 (OMIM 613593), all involved in immune-related pathways. All 59 genes were included in constructing the race-differentiated gene signature. A subset of 30 racial-differentiated genes were found significantly associated with overall breast cancer survival. We externally validated our 59-gene signature by fitting the model to TCGA data and evaluated the prediction performance of the gene signature in terms of both discrimination and calibration. The C-statistic was 0.81, indicating high discriminative ability in distinguishing AA and EA breast cancer patients. The Brier score, which measures disagreement between the observed outcome and a prediction, was 0.064, indicating high reliability and prediction accuracy. Conclusion: These findings provide insights into the biological differences in tumors and the survival disparity between AA and EA breast cancer patients. Citation Format: Jie Ping, Xingyi Guo, Fei Ye, Jirong Long, Loren Lipworth, Qiuyin Cai, William J. Blot, Xiao-ou Shu, Wei Zheng. Differences in gene-expression profiles in breast cancer between African and European-ancestry women [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 50.