Question: Kaplan Meier curve to validate a set of biomarkers
gravatar for ncl.lazzarini
4.9 years ago by
United Kingdom
ncl.lazzarini90 wrote:

I have a set of biomarkers (genes) and I'd like to validate it analysing the Kaplan Meier survival curve. I had a look at the literature and most of the time this curve is calculated using the expression value of a single gene. The patients are divided in two groups:  low expression and high expression, often using and arbitrary expression value threshold as the mean. Finally using those two group the curve is plotted. What is a common practice to do when the number of genes is higher than one? How the patients are assigned to the two groups?

Furthermore, is there a simple tool that allows me, given:

a) list of genes

b) expression values of those genes

c) survival information of the patients

to create/plot the Kaplan Meier curve?


ADD COMMENTlink modified 4.8 years ago by Biostar ♦♦ 20 • written 4.9 years ago by ncl.lazzarini90

I would suggest taking your gene set and doing an unsupervised clustering of your patients using gene expression profiles (e.g. K-means with k=2 or k=3). Then plot separate Kaplan Meier curves for your clusters and perform routine statistical tests for them. I guess that would be easy to do using a custom script in R.

ADD REPLYlink written 4.9 years ago by mikhail.shugay3.3k

But what would be the meaning of each group? Let's suppose k=2 and there is a significant difference between the 2 KM curves? What can I conclude? What is the value of such division in groups?

ADD REPLYlink written 4.9 years ago by ncl.lazzarini90

Anything of this kind that could predict clinical outcome is of a high value itself. The interpretation should be based on the up/down states of genes in those samples. I know that is quite a heuristic approach. Yet for example in breast cancer classification one has ER+, triple negative, etc - multiple subtypes based on combinations of expression of 3 receptor genes and it works fine.

ADD REPLYlink written 4.9 years ago by mikhail.shugay3.3k

You may want to checkout this website that allows you to easily plot KM plots (based on their data-sets)

ADD REPLYlink written 4.8 years ago by roy.granit800
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: 2886 users visited in the last hour