Hello everyone.
I have downloaded brain cortex normalized gene expression data from gtex portal. There are total 17609 genes whose expression is available. I want to specifically select those genes whose expression are significant in all the brain cortex and ignore other in my further TWAS study. Is there any way to do this?
Thank you.
This is how my dataset look like:
dput(expr_df[1:5,1:5])
structure(list(ENSG00000227232 = c(2.06593146369747, 0.592675922123552,
-0.109732708742409, 0.651749000024809, 0.636774336760061), ENSG00000268903 = c(1.19264174697961,
-0.507718909302338, -0.493924893351373, -1.327128982828, -0.347580384376274
), ENSG00000269981 = c(-0.347580384376274, -0.283623365386845,
-0.195936083375207, -0.794018835723402, 0.309052549457822), ENSG00000241860 = c(-1.05535688017356,
0.480224227652705, 1.14447383125234, 0.744935555245567, -0.89847561689514
), ENSG00000279457 = c(0.33468319748728, -0.493924893351373,
0.258376301003193, 0.507718909302338, 0.549701290113435)), row.names = c("GTEX-1117F",
"GTEX-111FC", "GTEX-1128S", "GTEX-117XS", "GTEX-1192X"), class = "data.frame")
•
written
by
rheab1230 •
1 answer
You will need a comprehensive set of RNA-seq data from the same batch and then do some sort of differential expression to enrich for genes with evidence to be specific (whatever this means) or preferentially expressed in this tissue. RNA-seq is inherently relative so with one tissue along you won't get far. From GTEx maybe you can download many tissues, and then use limma-trend (or something like this) to make such analysis. The simplest could be to perform a test where you contrast your brain tissue to the average of all other tissues and select those that are most significant.
Log in to answer this question.
Normalized how?
I got the normalized gene count from gtex portal. they use inverse quantile normalization
Can you explain the normalization or give me a link that describes it, please? Comparison across samples will always be caveated as RNA-seq yields relative measures.