I have a TCGA raw RNA-seq counts (HTseq) for tumor samples. I want to do survival analysis based on gene expression for a particular gene set.
I know that I can download normalized TCGA data with z-scores etc. But as I have a certain set of genes I would like to separate them based on the median expression of a gene - into "UP" (greater than the median) and "DOWN" (lower than the median).
In all similar tutorials or answers here, I see that people do differential analysis using DESeq or EdgeR and use built-in normalization. I do not need to do differential analysis, I just need to find associations between particular genes' expression and overall survival.
Can I use just some simple data normalization like CPM or log normalization on raw RNA-seq counts to do survival analysis as I described above?
•
written
by
fifty_fifty •
0 answers
No answers yet.
Log in to answer this question.
-
TPM RNA-seq data for differential expression analysis
written by Catalin-George •
-
RNA-seq analysis without replicates
written by fifty_fifty •
-
Do I have to separate my interest genes from my count matrix and then perform differential expressi…
written by Azade •
-
Correct Data from TCGA for EdgeR Analysis
written by joseph.landry •
-
Survival analysis based on a set of genes for TCGA data
written by fifty_fifty •
-
Converting CPM to FPKM or TMM?
written by klorilla •
-
Which RNA-seq unit should be used in machine learning models?
written by darklings
-
What is the cutoff used for define high or low expression level of gene for survival analysis
written by mygamefun3 •
-
edgeR normalization and partial correlation matrix calculation
written by moxu
-
Computing Normalized Counts For Exons In Edger
written by Nick
Hello,
Did you find an answer for this? Ideally any manual/solid giving solid guidelines? I am facing the same problem.
Thank you!
It's just edgeR or DESeq2 normalization. If you want to compare medians or any other metric then data have to be corrected for depth and composition -- exactly what mentioned tools do. Use the logcpms from edgeR or vst from DESeq2 and then do your median/quartile/whatever thingy on it. Don't overthink it.