Survival analysis for several genes

Question

[rogerbear]

Hi guys, I wonder how can I do survival analysis for several genes at the same time. I have the followup data of some patients(overall survival, disease-free survival), and i have the expression value of several genes (gene1, gene2, and gene3). To do survival analysis for one gene, I can divide the patients into two groups based on mean expression value of the gene and then analyze the survival data. But how can I analyze all the three genes at the same time ( I mean combine the expression of them, and consider them as one gene so I can group my patients). Any response would be appreciated. Thanks

Answer 1

[Kevin Blighe]

Hi,

You just need to construct a standard survival regression model of the form (in R):

library(survival)
Surv(Months, DFS) ~ gene1 + gene2 + gene3

NB - 'DFS' = disease-free survival

I have a tutorial here: Survival analysis with gene expression

###

For a quick example, try something like:

library(survminer)
ggsurvplot(survfit(Surv(Months, DFS) ~ gene1 + gene2 + gene3, data = survplotdata),
    data = survplotdata,
    risk.table = TRUE,
    pval = TRUE,
    break.time.by = 500,
    ggtheme = theme_minimal(),
    risk.table.y.text.col = TRUE,
    risk.table.y.text = FALSE)

Kevin

[rogerbear]

Hi Kevin, thanks for your reply. You tutorial helped me with my survival analysis with R. But you seems misunderstood my question. the survfit()~ gene1 + gene2 + gene3 feature enables me to analyze several genes at the same time and display all the genes on the same survival plot, but all the genes are still separately. what i want to achieve is to combine all the genes into one integrated gene, and then do survival analysis using this one "integrated gene" which contains information of all the three genes. I have the expression values of these three genes separately, how can i combine all the expression values?

[Kevin Blighe]

Hi, you will have to determine a way to summarise the expression of the 3 genes. You could try something like using the 'residuals' from a regression model of form ~ gene1 + gene2 + gene3, or some categorical encoding like High_High_High, Low_Low_Low, Other

[1769mkc]

Useful tutorial I have done the first where you fit multiple genes into cox regression model that part i know understand and did learn ,can you show some dummy data for this part "High_High_High, Low_Low_Low, Other" as this is neat way to summarized and test out group of genes which can be categorized based on gene class or enrichment which can be used

Answer 2

[huynguyen96.dnu]

The solution of @Kevin Blighe just performs a multivariate Cox regression model. If you want to perform an univariate Cox model, now you can do this task very easily with my recently developed tool named 'geneSA' (https://github.com/huynguyen250896/geneSA). Its output will automatically report genes statistically significant with survival outcome. Give it a try ;)

[seta]

as I see, both geneSA and RegParallel tools perform the similar analysis, univariate test as Kevin kindly explained here.

Dear Kevin, could you please explain a bit more about how we can consider multiple genes, say, only high expressed genes, as one single variable for cox regression (survival) analysis? It will be great if you can introduce any practical tutorials.

[GenoMax]

huynguyen96.dnu please create a single tools post instead of posting your tool in every past thread that refers to survival analysis.

[huynguyen96.dnu]

Oh I am sorry! It is my fault. What should I do. Should I remove these posts

[GenoMax]

That is fine. Just create a single tools post where you can provide information about where to find the tool, if there is publication, how to use the tool etc. It is tidier that way.