The overall goal of Project 4 is the development of a gene expressionbased predictor of outcome in prostate cancer. The hypothesis underlying this project is that the clinical heterogeneity of prostate cancer has a molecular signature, but that signature has yet to be identified. We will focus on the heterogeneity of outcome following radical prostatectomy of patients with organ-confined disease, and we will use DNA microarrays to identify the gene expression signature of propensity for tumor recurrence.
In Aim I, we will generate a high quality, clinically and pathologically annotated gene expression database of 150 primary tumors for which long-term clinical follow-up is available. RNAs generated from these frozen tumors will be subjected to DNA microarrays containing probes for 60,000 genes and ESTs.
In Aim II, we will use this gene expression database, together with clinical information, to generate a molecular predictor of recurrence. We will use supervised machine learning approaches to this problem, evaluating the accuracy of the outcome predictor by several criteria including random permutation testing, leave-one-out cross-validation testing, and testing on an independent dataset.
In Aim II of the project, we will work to develop a version of the outcome predictor that can be implemented in a more routine clinical setting, rather than in a highly specialized genomics research laboratory. To that end, we will explore extending the outcome prediction model to an immunohistochemical implementation, to a quantitative RT-PCR approach, and we will work to develop methods suitable for extracting this information from small, needle biopsy prostate specimens. At the time of completion of this project we aim to have developed a robust molecular predictor of outcome in prostate cancer, and to have translated this assay into one that can be evaluated by the wider community. It is anticipated that such molecular predictors of clinical behavior will be of value in individualizing treatment decision-making for patients with prostate cancer.
Showing the most recent 10 out of 261 publications