differential expression based on multiple biological repeats

Question

[Whetting]

Hi,

I downloaded the raw data from the current GTEX release (http://www.gtexportal.org/home/). I parsed the data to get a matrix of a couple of genes and their associated RPKM values in a bunch of different samples. E.g. I have 6 samples isolated from the cervix, 10 samples isolated from skin, 25 samples from vagina.

I am interested in calculating the differential expression of these genes in these tissues.

Since RPKM are not raw counts, I will not be able to use most standard pipelines to calculate differential expression. Could anyone recommend a tool that I could use? Does anyone have experience with DEGSeq?

Also, how do I combine samples? Can I just consider them to be replicates?

As always, thanks...

Answer 1

[Lorena Pantano]

Hi,

you can download count data from their site, but need to create an user first (http://www.gtexportal.org/home/datasets). After that, you should do DE with DESeq, edgeR or limma:voom to get the correct normalization. Then look for those genes.

It is not good to pick up a bunch of genes and do only DEG using RPKM, there is bias there for each sample, so there is no normalization between-samples.

And, yes I would consider those as replicated, everything that comes from the same tissue. Another issue is how to normalize since these are different tissues and the genes that can be DEG may be high. The bioConductor support site (https://support.bioconductor.org/local/search/page/) is a good way to get support for this. People tell that if the number of DEG is less than 30% is ok to do standard normalization inside those packages.

hope this help

Answer 2

[KVC_bioinfo]

LINK

Link above can be good starting point to understand the DESeq and edgeR