Abstract Background: Basal-like breast cancer (BLBC) have poor prognosis. Molecular similarities have been reported between BLBC and high grade serous ovarian cancer (HGSOC). To date, there have been no prognostic biomarkers specifically developed for BLBC or HGSOC that can provide risk stratification and inform treatment selections. In this study, we utilized RNA-seq data available from The Cancer Genome Atlas (TCGA) project to develop molecular signatures for risk stratification in BLBC, and further validated these signatures in HGSOC RNA-seq data from TCGA. Methods: Raw count of RNA-seq data were downloaded from TCGA for 190 BLBC and 374 HGSOC patients. The datasets were annotated with 56963 Ensembl gene IDs. Excluding 31375 gene IDs with no greater than 10 counts in at least 90% of the samples, totally 25228 unique Ensembl gene IDs were used. Progression-free interval (PFI) is the primary study endpoint. Analyses of differentially expressed genes were performed using 3 bioconductor packages: DESEq2, edgeR and voom/limma. Signatures based on commonly identified genes among the 3 analytic methods were established using weighted linear combination of gene expression levels. Their performance was evaluated in the BLBC and HGSOC datasets using Kaplan-Meier survival analysis with log-rank tests and Cox proportional hazard regressions. Results: Among 190 TNBC patients, 18 had recurrences within 2 years and 40 showed no recurrences for at least 5 years. These patients were used as recurrent vs. non-recurrent cases for differential expression analysis. 307 and 343 genes were differentially expressed based on adjusted p value threshold 0.05 and 0.01 in DESeq2 and edgeR analysis, respectively. voom/limma identified no genes differentially expressed based on adjusted p values, but 228 genes had unadjusted p values < 0.01 and were used in the following analysis. Taken together, 63, 58 and 21 genes were commonly identified by DESeq2/edgeR, DESeq2/limma and edgeR/limma analysis, respectively. All 3 signatures were able to significantly stratify the TNBC full dataset (n=190) using either 20-, 50- or 80-percentile as the cut-points. When evaluated in HGSOC patients using 80-percentile cut-point, both 63- and 58-gene signatures were able to significantly stratify patients into different risk groups (HR 2.16, 95% CI: 1.4-3.34, p < 0.001; HR 2.06, 95% CI: 1.36-3.11, p < 0.001). Multivariate Cox regression adjusting for age, grade and stage showed 63- and 58-gene signatures remained to be statistically significant in stratifying HGSOC patients (p = 0.0005 and 0.001, respectively). Conclusion: Gene signatures were specifically identified to prognosticate BLBC patients based on RNA-seq data from TCGA project. Which were able to classify HGSOC patients into differential risk groups. With further validations, these signatures may provide additional prognostic tools for clinicians to better manage triple-negative breast cancer that mostly overlap with BLBC, and HGSOC patients who are difficult-to-treat currently. Disclaimer The views expressed in this article are those of the authors and do not reflect the official policy of the department of Army/Navy/Air Force, Department of Defense, or U.S. government. Citation Format: Hu H, Zhang Y, Liu J, Raj-Kumar P-K, Yang H, Lee M, Kovatich AJ, Shriver CD. Development and validation of prognostic gene signatures for basal-like breast cancer and high grade serous ovarian cancer [abstract]. In: Proceedings of the 2018 San Antonio Breast Cancer Symposium; 2018 Dec 4-8; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2019;79(4 Suppl):Abstract nr P2-08-12.
Read full abstract