Despite progress in cancer therapy, early detection remains the most promising approach to improve long-term survival of patients with ovarian cancer. The insufficient performance of available single biomarkers and the observed heterogeneity among ovarian cancer patients strongly suggest the benefit of a multiplexed approach with complementary biomarkers to improve overall performance. Proteomic profiling involves the systematic analysis and characterization of all (or at least a large number of) proteins and their expression patterns in a given biological or clinical specimen. Bioinformatics provides the computational and statistical tools necessary for the analysis of the typically large volumes of profiling data. Recent advances in proteomic profiling technology have made it possible to apply bioinformatics methods to detect changes in protein expression patterns and their association to disease conditions, thereby hastening the identification of novel markers that may contribute to multi-marker combinations with a highly accurate diagnostic performance for the early detection of ovarian cancer. The success of such approaches, however, will require both statistically sound experiment design and biologically meaningful use of bioinformatics tools. In the observed profile data, true differences in protein expressions between samples from cancer patients and healthy controls are often overshadowed by many nondisease-related variances and biases. Among them, subtle differences in the so-called "pre-analytical variables" such as patient inclusion/exclusion criteria in study protocols, specimen collection, handling, and processing procedures, etc., could introduce systematic biases in the expression levels of some proteins that correlate with sample classifications. In general, it is difficult to detect such site-specific biases based on data from a single site alone.
The overall objective of this study is to use bioinformatics tools and proteomic profiles of serum samples collected from multiple institutions to discover and validate biomarkers for the early detection of epithelial ovarian cancer. A key feature of the study is the proposed experiment design in which multiple data sets from samples collected at independent and diverse sites are used specifically to alleviate the risk of false discovering from nondisease-related biases in data. Once the biomarkers are discovered, their protein identities will be determined. This will allow us to validate these biomarkers using additional assays and methods and to better understand their biological relevance. The Principal Investigator for many years has been working in the area of combining multiple tumor markers for the detection of stage I/II epithelial ovarian cancer. In several reports, such multivariate predictive models could reach a sensitivity > 70% for detection of early stage ovarian cancer at a specificity of 98% among apparently healthy women. With the relatively low prevalence of ovarian cancer, such a performance still falls slightly short for them to be applicable to a general population. With the discovery and validation of novel biomarkers for ovarian cancer from proteomic profiling, we should be able to further improve the performance of such models so that they could be used in a general population to screen for early stage ovarian cancer.
Based on results from our preliminary study, we fully expect that by the end of the proposed 2-year period, we will have a selection of biomarkers ready for further validation through large-scale clinical studies.