Analyzing Survival Time with Gene Expression Data
Author Information
Author(s): Wu Tongtong, Sun Wei, Yuan Shinsheng, Chen Chun-Hou, Li Ker-Chau
Primary Institution: University of Maryland, College Park, MD, USA
Hypothesis
Can a new method effectively analyze the relationship between gene expression profiles and survival time in patients?
Conclusion
The proposed method successfully identifies gene expression signatures that predict survival probabilities in breast cancer patients.
Supporting Evidence
- The method was tested on both simulated and real data.
- 22 gene expression profiles were identified as significantly correlated with survival rates.
- The log-rank test showed significant differences in survival rates between high and low risk groups.
Takeaway
Researchers created a new way to look at how gene activity relates to how long patients live, helping to find important gene patterns for predicting survival.
Methodology
A two-step procedure involving gene pre-selection using correlation or liquid association, followed by dimension reduction using modified sliced inverse regression for censored data.
Potential Biases
Potential bias due to high censoring rates in survival data.
Limitations
The method may miss important genes if the sample size is small, and the imputation method for survival probabilities may need improvement.
Participant Demographics
295 breast cancer patients, with a censoring rate of 73.2%.
Statistical Information
P-Value
2.2e-16
Confidence Interval
[0.313, 0.496]
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website