Tumor Proteomic Profiling Predicts the Resistance of Breast Cancer to Chemotherapy
D Shen, J He, DU Cheung, KF Faull, JP Whitelegge, RE Saxton, HR Chang. UCLA David Geffen School of Medicine, Los Angeles, CA; USC Keck School of Medicine, Los Angeles, CA
Background: Chemotherapy is widely used in breast cancer treatment, but the outcomes vary with some patients responding well and others responding poorly. We hypothesized that the profile of the differentially expressed proteins in tumor tissue may predict individual drug response.
Design: The SELDI-TOF mass spectrometric profiles of tumor tissues obtained from drug resistant and drug sensitive tumors were compared to identify the differences between the two. Fifty-two T2-T4 breast cancer tissues obtained prior to neoadjuvant chemotherapy were analyzed. Of these the first two thirds (35 cases) were allocated to a training set to select m/z peaks characteristic of resistant tumors. The candidate m/z peaks were used to develop a predicting rule to evaluate the remaining 17 specimens in the validation set.
Results: The proteomic peak differences were found most prominent between the drug-resistant breast tumors compared with those with various sensitivity by non-supervised hierarchical clustering. In the supervised classification, the KNN model with K=1 correctly classified 100% of resistant tumors (4/4), and 84.6% of the tumors with favorable response (11/13) with an accuracy rate of 92.3% in the validation set. Furthermore, a single peak at m/z 16,906 correctly separated 88.9% of the tumors with pathologically complete response, and 91.7% of the resistant tumors in the entire group.
Conclusions: The data suggests that breast cancer protein biomarker profiling may be used to pre-select patients for optimal treatment.
Monday, March 9, 2009 9:30 AM
Poster Session I Stowell-Orbison/Autopsy Award # 47, Monday Morning