Related Experiment Videos
Accurate molecular classification of human cancers based on gene expression using a simple classifier with a
Kerby A Shedden1, Jeremy M G Taylor, Thomas J Giordano
1Department of Statistics, University of Michigan, Ann Arbor, MI 48109-1027, USA. kshedden@umich.edu
The American Journal of Pathology
|October 28, 2003
Summary
A simple gene expression algorithm accurately classifies human tumors, mimicking pathologist strategies. This approach requires fewer genes than complex methods for robust cancer diagnosis.
Area of Science:
- Biotechnology
- Bioinformatics
- Oncology
Background:
- Gene expression profiling is used for cancer classification.
- Sophisticated statistical learning methods are currently employed.
Purpose of the Study:
- To develop a simpler, accurate tumor classification strategy.
- To investigate a tree-based framework using limited pathological information.
Main Methods:
- Utilized a simple, tree-based algorithm for gene expression analysis.
- Classified human tumors based on gene expression levels.
- Compared performance against sophisticated classifiers using thousands of genes.
Main Results:
- A small number of genes (45) achieved high accuracy in tumor classification.
- Correctly classified 157 out of 190 malignant tumors.
- Demonstrated robustness across different labs, platforms, and metastatic vs. primary tumors.
Conclusions:
- A straightforward gene expression classification strategy can be as accurate as complex methods.
- Mimicking surgical pathologist classification strategies is effective for cancer diagnosis.
- This approach offers robust and accurate cancer diagnosis from gene expression profiles.