Gene Expression Predictors of Breast Cancer Outcomes
Abstract
The integration of currently accepted risk factors with genomic data carries the promise of focusing the practice of medicine on the individual patient. Such integration requires interpreting the complex, multivariate patterns in gene expression data, and evaluating their capacity to improve clinical predictions. We do this here, in a study of predicting nodal metastatic states and relapse for breast cancer patients.
DNA microarray data from samples of primary breast tumors were analyzed using non-linear statistical analyses to evaluate multiple patterns of interactions of groups of genes that have predictive value, at the individual patient level, with respect to lymph node metastasis and cancer recurrence.
We identify aggregate patterns of gene expression (metagenes) that associate with lymph node status and recurrence, and that are capable of honestly predicting outcomes in individual patients with about 90% accuracy. The identified metagenes define distinct groups of genes, suggesting different biological processes underlying these two characteristics of breast cancer. Initial external validation comes from similarly accurate predictions of nodal status of a small sample in a quite distinct population group.
Multiple aggregate measures of gene expression profiles define valuable predictive associations with lymph node metastasis and disease recurrence for the individual patient. These results indicate the potential for gene expression data to aid in achieving more accurate individualized prognosis. Importantly, this is evaluated in terms of precise numerical predictions, via ranges of probabilities of outcome, for the indidividual patient. Such precise and statisticallly valid assessments of patient-specific risk will ultimately be of most value to clinical practitioners faced with treatment decisions.
Authors
Erich Huang, Skye H Cheng, Holly Dressman, Jennifer Pittman, Mei-Hua Tsou, Cheng-Fang Horng, Andrea Bild Edwin S. Iversen, Ming Liao, Chii-Ming Chen, Mike West, Joseph R Nevins and Andrew T Huang
Contact
- Erich Huang (huang008@mc.duke.edu)