RPKM in GTEx and TCGA

[wbb1813]

Hi guys,

You know, GTEx level 3 data is rpkm value, and TCGA data is RSEM and part have rpkm.

Now I want compare TCGA data with GTEx data, so I use TCGA rpkm vs GTEx rpkm, unfortunately I get anout 15,000 up regulate genes. When I check the two data, I find TCGA rpkm value is bigger than GTEx value.

Questions is :
1.Both TCGA rpkm and GTE rpkm are rpkm value but Why are they different?
2.Some subtypes in TCGA only have RSEM value, if I want to compare these subtypes with GTEx data, are there some methods to solve this problems?

[wbb1813]

Is there anyone meet the same problem?

[Tao Sheng]

Hi just a curious how you do the compare.

Since I downloaded the gene RPKM file from http://www.gtexportal.org/home/datasets2

I found the gene ids are Ensembl ids and in total there are 55994 rows, and TCGA gene RPKM files have 20553 rows, responding to the gene names.

Did you do the conversion? and how did you deal with the situation that several Ensembl ids respond to 1 gene name, you calculate the average or add them up???

[llltrunkslll]

Hi guys,

You know, GTEx level 3 data is rpkm value, and TCGA data is RSEM and part have rpkm.

Now I want compare TCGA data with GTEx data, so I use TCGA rpkm vs GTEx rpkm, unfortunately I get anout 15,000 up regulate genes. When I check the two data, I find TCGA rpkm value is bigger than GTEx value.

Questions is :
1.Both TCGA rpkm and GTE rpkm are rpkm value but Why are they different?
2.Some subtypes in TCGA only have RSEM value, if I want to compare these subtypes with GTEx data, are there some methods to solve this problems?

As far as I know, you are not able to directly compare RSEM (which by some is considered the superior metric) directly to RPKM. The TCGA and GTEx assemblies are being done against different genomes (i.e. Refseq or UCSC genome builds). However you can make some indirect comparisons by normalizing to tissues common between the GTEx ones and the normal tissue samples from TCGA. This isn't perfect by you can begin to speak to relative differences. Hope that helps.