Will Kallisto show a decrease in the TPM value if I have knocked out 2 of the 8 exons from a gene?
48 views
Skip to first unread message
Simoni Tiano
Nov 29, 2021, 2:09:45 PM11/29/21
to kallisto and applications
Hello,
I was wondering if Kallisto will show a decrease in the TPM value if I have knocked out 2 of the 8 exons from a gene? I have confirmed the knock out using qPCR however, when I align with Kallisto the TPM values seems similar across my wild type and knockout samples. Would Kallisto identify the 6 exons as a gene?
Thank you.
Jim Notwell
Dec 28, 2021, 10:08:15 PM12/28/21
to kallisto and applications
Hello,
To answer your question, you must first consider what the transcripts per million (TPM) values are capturing: counts/reads per base, corrected for the total number of fragments sequenced. Harold Pimentel has a helpful blog post explaining the details here.
In your experiment, the exon deletion should result in fewer relative reads originating from the locus. Kallisto will still pseudoalign the reads for quantification. If you've updated your GTF/index to include a 6-exon transcript, you will have a smaller numerator (counts/reads) and smaller denominator (transcript length), resulting in similar TPM values.
Even if you have not updated your GTF/index, you may still see similar TPM values. I encountered a similar situation in grad school. Reads/counts originating from the remaining exons dwarfed the loss in signal from an exon deletion.
Here, it's important to remember that Kallisto quantifies expression at the transcript level (transcript TPMs corresponding to the same gene can be summed to produce gene TPMs). If you include two transcripts in your GTF/index - the 6 exon form and the 8 exon form - you should see obvious differences in the transcript-level quantifications of your samples, even if you don't observe them at the gene-level quantification.
It sounds like you've already confirmed your KO, but a way to do this with RNA-seq data is to inspect the read pile-ups (wiggle files) in a genome browser (for example, using the --outWigType and --outWigStrand argument in STAR).
Good luck,
Jim
To answer your question, you must first consider what the transcripts per million (TPM) values are capturing: counts/reads per base, corrected for the total number of fragments sequenced. Harold Pimentel has a helpful blog post explaining the details here.
In your experiment, the exon deletion should result in fewer relative reads originating from the locus. Kallisto will still pseudoalign the reads for quantification. If you've updated your GTF/index to include a 6-exon transcript, you will have a smaller numerator (counts/reads) and smaller denominator (transcript length), resulting in similar TPM values.
Even if you have not updated your GTF/index, you may still see similar TPM values. I encountered a similar situation in grad school. Reads/counts originating from the remaining exons dwarfed the loss in signal from an exon deletion.
Here, it's important to remember that Kallisto quantifies expression at the transcript level (transcript TPMs corresponding to the same gene can be summed to produce gene TPMs). If you include two transcripts in your GTF/index - the 6 exon form and the 8 exon form - you should see obvious differences in the transcript-level quantifications of your samples, even if you don't observe them at the gene-level quantification.
It sounds like you've already confirmed your KO, but a way to do this with RNA-seq data is to inspect the read pile-ups (wiggle files) in a genome browser (for example, using the --outWigType and --outWigStrand argument in STAR).
Good luck,
Jim
Reply all
Reply to author
Forward
0 new messages