Abstract P4-05-23: Validation of proteome based molecular stratification of breast cancer using 1,229 patient samples from four independent datasets

Abstract

Abstract Introduction. In routine diagnostics, four genes are used to determine the molecular subtype and the optimal systemic therapy for breast cancer patients, ESR1, PGR, HER2, and in some cases MKI67. While these can be measured by different molecular genetic methods, immunohistochemistry is still the most widely used technique. The aim of present study was to determine how effective recent proteomic technologies are in measuring the expression of these genes. Methods. A literature search was performed to identify available proteome-level datasets. Differential expression was compared using Mann-Whitney test. Receiver operating characteristics (ROC) was computed to validate discriminative power and to determine the optimal cutoff values. Each dataset was processed separately. Results. All together four datasets with a total of 1,229 specimens with proteome level-data were identified. One dataset was generated using RPPA (reverse phase protein lysate microarray, n=874) and three using LC-MS (liquid chromatography-mass spectrometry, n=355). In all four datasets combined, a total of 7,419 proteins were measured. It was not possible to analyze all combinations because of limited data availability - HER2 data was available in the RPPA study only, and ESR1/PGR in the LC-MS studies only. The determination of HER2 status by IHC and HER2 protein expression by RPPA show very high correlation (p=2.9E-18). ESR1 protein expression determined by LC-MS show a significant correlation to IHC results (p=0.04). The ROC analysis for HER2 delivered an area under the ROC curve (AUC) of 0.74 (p=1.9e-20). For ESR1, the AUC was 0.61 (p=0.03). ESR1 and PGR show a moderate correlation in LC-MS data (correlation coefficient=0.17, p=0.04). Finally, when comparing MKI-67 expression between normal (n=53) and tumor samples (n=65) within one dataset, tumor samples had significantly higher expression (fold change=2.22, p=3.9E-05). Conclusions. The RPPA method achieved high correlation with IHC results. Our results suggest that proteomic technologies will be capable to deliver molecular stratification enabling the discrimination of breast cancer subtypes in the future. Citation Format: Balazs Gyorffy. Validation of proteome based molecular stratification of breast cancer using 1,229 patient samples from four independent datasets [abstract]. In: Proceedings of the 2019 San Antonio Breast Cancer Symposium; 2019 Dec 10-14; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2020;80(4 Suppl):Abstract nr P4-05-23.