Following the analysis of survival and prediction by other clinical traits here I used the same survival object to investigate if there are any CpGs which status is predictive of survival (days to death). Made sure that I have the same number of patients in DNA methylation data sets and in the survival object.
Left panel is for mb and the right panel is for mbc. Obviously, the signal is not strong
I also performed permutations to calculate FDR, for mb. FDR is way too high.
P value |
FDR, % |
Number of targets |
|---|---|---|
0.05 |
68.91 |
1776 |
0.01 |
56.04 |
430 |
0.005 |
55.94 |
217 |
0.001 |
52.29 |
48 |
Similar calculations for mbc. This seems even worse.
P value |
FDR, % |
Number of targets |
|---|---|---|
0.05 |
79.9 |
1721 |
0.01 |
70.18 |
386 |
0.005 |
65.23 |
212 |
0.001 |
62.7 |
48 |
In the Sweave file I also performed analyses for the progression free interval but I didn't include the information for the days to the last follow-up and it might be as relevant here as for the days to death. Need to re-do this analysis.