Hello, I have a RNAseq data(human) of 9 samples with 3 replicates each. the mapping to the reference was performed by STAR and quantification was done by RSEM. The differential expression study on genes was performed by DESeq2(taking the raw counts), but it failed for isoforms.
1) Is there some other tool with which we can perform a DE study on isoforms? 2) can we use the TPM counts instead of the raw counts for performing the DE analysis? if no. then why? 3) using tximport + deseq2 on TPM values, will this work for DE isoform study?
•
written
by
glady
2 answers
To be honest, I've typically parsed the counts from Salmon / Sailfish directly (when I tried RSEM, the alignment either took a long time, or the quantification didn't seem as good: however, I assume that you have had better luck...).
However, if you are worried about bugs in your custom code (because something seems strange in DESeq2), there is a package that is supposed to help with import to DESeq2:
https://bioconductor.org/packages/release/bioc/html/tximport.html
I haven't tested it personally, but hopefully you find it useful!
It is supposed to work with RSEM output: https://bioconductor.org/packages/release/bioc/vignettes/tximport/inst/doc/tximport.html#rsem
Log in to answer this question.
-
Current best practice for alignment and quantification for downstream analysis in DESeq2
written by yura.grabovska •
-
What purposes can TPM values be used for?
written by JH •
-
To get p-values for the TPM
written by VITALA •
-
Differential Expression Analysis on RNAseq - RSEM expected_count from UCSC RNA-Seq toil recompute
written by Bioinformatician •
-
RSEM isoform.results outputs for edgeR
written by weichengz •
-
Compare expression level of isoforms among conditions - TPM or deseq2 normalised counts?
written by tianshenbio
-
TERF2 shRNA silencing is visible in WB and qPCR but not reflected in PE RNASeq
written by rohitsatyam102
-
Which normalization method to use FPKM/TPM?
written by glady
-
RPKM values for mapped miRNA reads
written by glady
-
Selecting the consensus/overlapped genes from the DE study
written by glady
What do you mean with DESeq2 "failed" on raw isoform counts? It sounds like the correct approach to me. RSEM does not return integer isoform counts however, there will be some rounding up to do...
RSEM gives you two files as output - "filename.genes.results" & "filename.isoforms.results. the second file contains the expected_counts on isoforms as well as the TPM counts.
These expected counts are not integers right? The only issue I see with using DESeq2 with it is to rounding them up. What else didn't work in your case?
Yes, the counts are in integers. But if you use these counts as inputs to DESeq2, it gives you false results. Because the DESeq2 works very well for the gene counts and not for isoform counts, since the isoform are usually overlaps in the genes.