Use of Proteomic Patterns in Serum to Identify Ovarian Cancer

In this study, a new high-order bioinformatics tool used to identify differences in proteomic patterns in serum was evaluated for its ability to detect the presence of cancer in the ovary. The proteomic pattern is generated using matrix-assisted laser desorption and ionization time-of-flight and surface-enhanced laser desorption and ionization time-of-flight mass spectroscopy from thousands of low-molecular-weight serum proteins. Proteomic spectra patterns were generated from 50 women with and 50 women without ovarian cancer and analyzed on the Protein Biology System 2 SELDI-TOF mass spectrometer (Ciphergen Biosystems, Freemont, CA) to find a pattern unique to ovarian cancer. In the graph of the analysis, each proteomic spectrum is comprised of 15,200 mass/charge (m/z) values located along the x axis with corresponding amplitude values along the y axis. By comparing the proteomic spectra derived from the serum of patients with known ovarian cancer to that of disease-free patients, a profile of ovarian cancer was identified in the peak amplitude values along the horizontal axis. The comparison was conducted using repetitive analysis of ever smaller subsets until discriminatory values from five protein peaks were isolated. The validity of this pattern was tested using an additional 116 masked serum samples from 50 women known to have ovarian cancer and 66 nonaffected women. All of the subjects with cancer and most of the women with no cancer were from the National Ovarian Cancer Early Detection Program at Northwestern University. The nonaffected women had been diagnosed with a variety of benign gynecologic conditions after evaluation for possible ovarian cancer and were considered to be a high-risk population. Serum samples were collected before examination, diagnosis, or treatment and frozen in liquid nitrogen. The samples were thawed and added to a C16 hydrophobic interaction protein chip for analysis. In the validation set, 63 of the 66 women with benign ovarian conditions were correctly identified in the spectra analysis. All 50 patients with a diagnosis of ovarian cancer were correctly identified in the analysis, including 18 women with stage I disease. Thus, the ability of proteomic patterns to detect the presence of ovarian cancer had a sensitively of 100%, a specificity of 95%, and a positive predictive value of 94%. In comparison, the positive predictive value for serum cancer antigen 125 in the set of patients was 35%. Additionally, no matching patterns were seen in serum samples from 266 men with benign and malignant prostate disease.