ssGSEA on RNA-Seq data from TCGA

Question

[oriolebaltimore]

Hi:

Given TCGA RNA-Seq data (level 3), how one can do single sample GSEA?

For example, TCGA RNA-Seq data has

Gene      Raw counts     RPKM
ALK       434            2.3
..        ..             ..

Can we create a GCT file of Raw counts or RPKM to ssGSEA in gene pattern?

Thanks
Adrian

[orzech_mag]

Hi,

ssGSEA in comparison with GSEA calculates separate enrichment scores for each pairing of a sample and gene set.

However, both need the .GCT files as an input. From the TCGA you need to download the level 3 data, however it has to be expression. So, after some preprocessing your .GCT should look like this:

#1.2
no. of rows       no. of columns
NAME Description Sample1 Sample2 ...
gene1 NA               exp. value
gene2 NA
...

I think that there is no possibility to calculate those statistics from raw counts and rpkm. Raw counts may be biased, so I think that you should consider some normalisation step for such analysis to make it comparable with other studies.

Best regards!

[fuyingxue]

Could anyone explain how ssGSEA process the gene expression data to rank the gene for each patient sample? Dose it need to be compared with the normal sample?

Answer 1

[John Ma]

I know this is an old thread, but SSGSEA can be calculated using the Bioconductor package GSVA. If you use RPKM, use ssgsea <- gsva (RPKM, method="ssgsea", kcdf="Gaussian", ...); if you use raw counts, use ssgsea <- gsva (counts, method="ssgsea", kcdf="Poisson", ...)

[igor]

I think kcdf parameter is only applicable when method = "gsva". According to the docs:

Character string denoting the kernel to use during the non-parametric estimation of the cumulative distribution function of expression levels across samples when method="gsva".