Comparing Gene Selection Techniques for Cancer Classification
Author Information
Author(s): Lai Carmen, Reinders Marcel JT, van't Veer Laura J, Wessels Lodewyk FA
Primary Institution: Delft University of Technology
Hypothesis
Univariate gene selection techniques may perform better than multivariate techniques in classifying cancer datasets.
Conclusion
Univariate selection approaches yield better results than multivariate approaches in most cancer datasets studied.
Supporting Evidence
- Univariate approaches performed better in five out of seven datasets.
- The Top Scoring method was the best multivariate approach in two datasets.
- Complex multivariate methods may overfit due to small sample sizes.
Takeaway
This study looked at different ways to pick important genes for cancer classification and found that simpler methods often work better than more complex ones.
Methodology
The study compared univariate and multivariate gene selection techniques using an unbiased protocol across seven cancer gene expression datasets.
Potential Biases
The training and validation phases were not strictly separated, leading to potential information leaks.
Limitations
The results may not be generalizable due to the small number of datasets and potential biases in previous studies.
Participant Demographics
The datasets included various cancer types and were measured using different technologies.
Statistical Information
P-Value
p<0.05
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website