Question: What is the cutoff used for define high or low expression level of gene for survival analysis
gravatar for mygamefun3
8 months ago by
mygamefun30 wrote:

Hi everyone

In RNA-seq analysis, we need to separate samples into two groups for survival analysis. How can I define high level or low level for a gene according to counts or FPKM. Use median? average or quantile?

In TCGA or Oncomine, how are they define the cutoff for a gene ?


rna-seq • 791 views
ADD COMMENTlink written 8 months ago by mygamefun30

There's no definitive answer to your question. I would not advise going by the median or average. Quantile is a reasonable idea, or tertiles, with the higher third being regarded as "high expression".

An even better idea would be to convert your data to the Z scale, i.e., standard deviations from the mean, and then choose absolute 3, 4, 5, or 6 (3, 4, 5, or 6 standard deviations from the mean) as potential cut-offs. I trust that you have QC'd your data already and that low count transcripts have been removed.

ADD REPLYlink modified 8 months ago • written 8 months ago by Kevin Blighe28k

Thanks for your advice.

ADD REPLYlink written 8 months ago by mygamefun30

Hi Kevin,

I would like to look at survival of a specific gene between high and low expression tumor samples. For this, I want to divide tumor samples into high and low. I have counts data and transformed them into z-score.

Do you think the below one is right way to divide the samples?

event_rna <- t(apply(z_rna, 1, function(x) ifelse(x > 1.96,1,ifelse(x < -1.96,2,0))))

Or Should I follow this high and low samples separation

In that they used FPKM and took the median and based on that they separated samples into high and low.

Which one is the right solution?

ADD REPLYlink modified 6 weeks ago • written 6 weeks ago by Biologist70

Hey bro, There is no right or wrong. Your function looks okay!

ADD REPLYlink written 6 weeks ago by Kevin Blighe28k

Thank you very much Kevin.

ADD REPLYlink written 6 weeks ago by Biologist70
Please log in to add an answer.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 1404 users visited in the last hour