Hi, I have a question of using sample Data.
I want to try some experiments of using classification method for cancer classification using DNA methylation values.
I will use methylation beta value of TCGA solid tumor for cancer data.
A question I have is that I want to use normal sample data to compare the methylation value of specific CpG sites.
Here, I am confused in whether I should use healthy control sample data, or whether I can use normal tissue sample of patients who have cancer.
For example, I have methylation beta value of cpg00001 in solid tumor tissue of breast cancer patient. And to compare this, I have cpg00001 value in solid tissue normal of same breast cancer patient.
Is it possible to use like this?
•
written
by
miniymay101 •
Recent trends suggest that methylation patterns are inherited, proliferate within the individual and are subject to modification by environmental conditions. With that in mind, it's probably better to control with an healthy tissue of the same patient. If such is not available (as is the case in many TCGA cases) perhaps match demographic characteristics (e.g. ethnic, gender) between patient and control individual.
Umm.. I cannot 100% understand "control with an healthy tissue of the same patient".. Does this mean that I can use normal tissue of the same patients as normal one and proceed cancer classification(TCGA tumor v.s normal) based on this?
Log in to answer this question.
How are you sure you tissue is normal part ? What do you use to distinguish normal and tumoral tissue ? Is it the same organ ? How can you be sure it's the same part of the tissue you take ?
At the end best think will be to compare to an other normal patient to check if in the same tissue there is no statistic differences between samples in normal tissue.
I can not sure 100% the tissue is normal part, but I will gonna use TCGA data set and It says there are data sets which are normal samples, so I thought I could maybe use them as normal one. Isn't TCGA data authorized one?
Ok so you bet that you will not find tumoral mutation in your control , isn't it ? I think you will find some anyway ...