Serum proteomic profiling distinguishes estrogen receptor (ER) positive and ER negative breast cancers

Abstract

578 Background: ER (+) and ER (-) breast cancers have different clinical response to current therapies and distinct clinical outcomes. Identification of distinguishable serum protein profiles of these two breast cancer entities may facilitate biomarker discovery for disease diagnosis and elucidation of potential therapeutic targets. Method: Serum samples of 30 patients with ER (+) and 11 patients with ER (-) breast cancer were divided 70:30 into training and testing sets. Patients in both training and testing sets were frequency matched by age and race. Another independent sample set of 48 white postmenopausal women with ER (+) breast cancer was used to validate the finding. Samples were analyzed in duplicate on Ciphergen IMAC3-Cu ProteinChip arrays with PBSIIc SELDI-TOF-MS. Spectral preprocessing with Ciphergen ProteinChip and CiphergenExpress software included baseline subtraction, normalization to total ion current, and internal peak alignment. Averaged peak intensities of over 180 potential features (peaks) were submitted to Ciphergen Biomarker Patterns™ software (V.5) for scripted generation of multiple classification trees (CART). Two hundred and ten decision trees were generated for the two-group comparison using the training set. The performance of the best and top 10% classifiers from the training, testing, and validation set is reported. Conclusions: There are features in serum protein profiles that distinguish ER (+) from ER (-) breast cancer patients with a relatively high accuracy. An independent sample validation set with both ER (+) and ER (-) breast cancers is needed to further confirm this finding. A more specific serum protein profile needs to be identified for each group when they are compared with non-malignant sample sets. [Supported by DoD IDEA grant BC 021912 to AEL] No significant financial relationships to disclose.