Question: Differential expression tool that can take TPM as input?
0
gravatar for Tye Kahn
9 months ago by
Tye Kahn10
Tye Kahn10 wrote:

Due to my analysis, I need to use TPM as input for my differential expression analysis.

I tried to use both DESeq2 and EdgeR by skipping both normalization and independent filtering steps. Even so, the number of DEGs is 3 times lower than using the original raw data (in DESeq2).

Does any of you know of any tool that allows using TPM as input for this kind of analysis?

ADD COMMENTlink modified 9 months ago by ATpoint38k • written 9 months ago by Tye Kahn10
3
gravatar for ATpoint
9 months ago by
ATpoint38k
Germany
ATpoint38k wrote:

https://support.bioconductor.org/p/126817/

Please also use the search function and google as this has been addressed literally dozens of times.

ADD COMMENTlink written 9 months ago by ATpoint38k

Already searched for it, but none of the results was informative. About the link, CPM, unlike TPM, does not take gene length into account. Do you think that may have any effect on the results? I would think so.

ADD REPLYlink written 9 months ago by Tye Kahn10
1

None of it is ideal. You should start from raw counts if possible. If not and TPM is the only thing you have then what does it matter, results will anyway be unreliable, so interpret them with care and try to validate the important genes with either experiments or published RNA-seq data in a similar setup. Gene length correction down-penalizes counts of long genes and by this reduces its power, this is (from what I understand) why it is not by default done in tools like edgeR or DESeq2. But as said, if TPM is what you have with no alternative, go and try the pipeline suggested at BioC and then see if you get anything out of it. Maybe you already have genes from which you know that they must / should change in your setup, and some that should not, so you can get an idea how robust the results are. You should also think about how drastic you expect changes to be. TPM cannot correct for library composition (see e.g. on Youtube the StatQuest videos on DESeq2 and edgeR library normalization to learn what this is and why it is important to correct for it), so if you expect major changes in the transcriptional landscape like mostly genes being upregulated or vice verse, your TPM counts might be skewed. Again, be careful with the results and try to confirm important genes that you want to build on.

ADD REPLYlink modified 9 months ago • written 9 months ago by ATpoint38k

the length normalisation does not make to much of a difference as it is hopefully the same in all samples - and you only compare the sample gene between samples. Meaning the results you got for CPM should be usefull.

ADD REPLYlink written 9 months ago by kristoffer.vittingseerup3.4k
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 574 users visited in the last hour