Identifying genomic signatures for predicting breast cancer outcomes.

详细信息   

• 作者：Rathnagiriswaran ; Shruti.
• 学历：Master
• 年：2008
• 导师：Guo, Nancy Lan,eadvisorCukic, Bojan,eadvisor
• 毕业院校：West Virginia University
• ISBN：9781109420616
• CBH：1471364
• Country：USA
• 语种：English
• FileSize：2944598
• Pages：94

文摘

Predicting the risk for recurrence in breast cancer patients is a critical task in clinics. Recent developments in DNA microarrays have fostered tremendous advances in molecular diagnosis and prognosis of breast cancer. The first part of our study was based on a novel approach of considering the level of genomic instability as one of the most powerful predictors of clinical outcome. A systematic technique was presented to explore whether there is a linkage between the degree of genomic instability, gene expression patterns, and clinical outcomes by considering the following hypotheses； first, the degree of genomic instability is reflected by an aneuploidy-specific gene signature； second, this signature is robust and allows breast cancer prediction of clinical outcomes. The first hypothesis was tested by gene expression profiling of 48 breast tumors with varying degrees of genomic instability. A supervised machine learning approach of employing a combination of feature selection algorithms was used to identify a 12-gene genomic instability signature from a set of 7657 genes. The second hypothesis was tested by performing patient stratification on published breast cancer datasets using the genomic instability signature. The results concluded that patients with genomically stable breast carcinomas had considerably longer disease-free survival times compared to those with genomically unstable tumors. The gene signature generated significant patient stratification with distinct relapse-free and overall survival log-rank tests； p <； 0.05； n ＝ 469). It was independent of clinical-pathological parameters and provided additional prognostic information within sub-groups defined by each of them. The importance of selecting patients at high risk for recurrence for more aggressive therapy was realized in the second part of the study, considering the fact that breast cancer patients with advanced stages receive chemotherapy, but only half of them benefit from it. The FDA recently approved the first gene test for cancer； MammaPrint, for node-negative primary breast cancer. Oncotype DX is a commercially available gene test for tamoxifen-treated, node-negative, and estrogen receptor-positive breast cancer. These signatures are specific for early stage breast cancers. A population-based approach to the molecular prognosis of breast cancer is needed for more rational therapy for breast cancer patients. A 28-gene expression signature was identified in our previous study using a population-based approach. Using this signature, a patient-stratification scheme was developed by employing the nearest centroid classification algorithm. It generated a significant stratification with distinct relapse-free survival log-rank tests； p <； 0.05； n ＝ 1337) and overall survival log-rank tests； p <； 0.05； n ＝ 806), based on the transcriptional profiles that were produced on a diverse range of microarray platforms. This molecular classification scheme could enable physicians to make treatment decisions based on specific characteristics of patients and their tumor, rather than population statistics. It could further refine subgroups defined by traditional clinical-pathological parameters into prognostic risk groups. It was unclear, whether a common gene set could predict a poor outcome in breast and ovarian cancer, the most common malignancies in women. The 28-gene signature generated significant prognostic categorization in ovarian cancers log-rank tests； p <； 0.0001； n ＝ 124), thus, confirming the clinical applicability of the gene signature to predict breast and ovarian cancer recurrence.