|
|
|
|
Building a Predictive Model Using Array Technologies with an Application to Breast Cancer Lue Ping Zhao and William
Whipple Neely The process of translating microarray technologies from bench-side to bed-side clinical services has started. One of the key translations of this basic science is to use genomic data to build predictive models for assessing disease diagnosis and prognosis. To date there have been several proof-of-concept applications of gene expression data to clinically relevant outcomes, including classification of cancers, prediction of metastasis and prognosis following surgery. While preliminary results are encouraging, genomic technologies have yet to produce any major diagnostic agents as documented in a report produced by an FDA working group. Furthermore, while the idea of using gene expression data to construct diagnostic tests appears to be straightforward, there are numerous challenges to building such models. Many of these challenges come from the central problem of how to construct and validate diagnostic or predictive models with gene expression data. In this paper we propose a general framework for developing and critiquing predictive models that may be used for diagnosis. It is our hope that this framework will address some of the key challenges to developing diagnostic and prognostic tests that incorporate genomic data. This framework is illustrated with an application to breast cancer. |
|
Graybill Conference |
