Kernel Methods for Accelerated Failure Time Model with High-Dimensional Data
Author Information
Author(s): Liu Zhenqiu, Chen Dechang, Tan Ming, Jiang Feng, Gartenhaus Ronald B
Primary Institution: University of Maryland Greenebaum Cancer Center
Hypothesis
Can kernel-based methods improve the efficiency of variable selection in accelerated failure time models for ultra-high dimensional genomic data?
Conclusion
The proposed methods can effectively identify survival-associated prognostic factors and predict survival outcomes using ultra-high dimensional genomic data.
Supporting Evidence
- The proposed method performs well with limited computational studies.
- AKRR can correctly identify survival-associated variables with high accuracy.
- The test RRMSE is lower with the proposed method compared to traditional LASSO methods.
- Simulation studies show that the proposed methods outperform existing methods in variable selection.
Takeaway
This study shows a new way to find important genes that help predict how long patients with cancer might live, even when there are a lot of genes to look at.
Methodology
The study used a nonlinear kernel ridge regression model and adaptive kernel ridge regression for variable selection in survival analysis.
Potential Biases
The two-step procedure in current microarray analysis may introduce bias and inaccuracies.
Limitations
The study may not generalize to all types of genomic data or other survival analysis contexts.
Participant Demographics
The median age of participants was 63 years, with 56% being men.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website