0
$\begingroup$

I want to look at an RNA-seq data-set on 1 condition to look for expressed genes. As there is only one condition, I can’t perform differential analysis.

The only thing I can think to do is set a threshold (i.e. all replicates must have > 10 gene counts) then say these are expressed. But I don’t think this is a very good way of doing it.

Do you know of any better ways that we can do this?

$\endgroup$
0

3 Answers 3

2
$\begingroup$

If you've got enough replicates (ideally at least 6, but can be done with 3), you should be able to do a Variance-stabilised Transformation using DESeq2:

vsd <- vst(dds, blind=FALSE)
rld <- rlog(dds, blind=FALSE)
head(assay(vsd), 3)

More details on this can be found in the DESeq2 Vignette:

https://bioconductor.org/packages/release/bioc/vignettes/DESeq2/inst/doc/DESeq2.html#extracting-transformed-values

$\endgroup$
1
$\begingroup$

Answer from @james-hawley, converted from comment:

In any high throughput sequencing experiment, there will be lowly expressed RNA molecules that are not captured by the library prep + sequencing. So defining what "is expressed" is somewhat arbitrary from RNA-seq data, regardless of how many replicates you have and what method you use. That said, you can control some of the noise in your RNA-seq data using a variance-stabilized transformation, like @gringer suggests, and look at the distribution of genes/transcripts. Some threshold like $rlog \ge x$ in at least $n$ samples is usually a good starting point (for some arbitrary $x$ and $n$).

$\endgroup$
1
$\begingroup$

I don't think there is a very good way to do what you want to do. I'd think some kind of RPM would be better than strictly raw counts; you have to account for overall library size. Or TPM might be better.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.