from row count to tpm

Question

[michelafrancesconi9]

Hello, I have a raw count matrix and I did some analysis with DESeq2 but now I want a TPM normalized matrix for CIBERSORTx. I have the fastq files and I read some post that say that I have to use Kallisto but I do not understand how I made this type of matrix and how I can find the gene length. Can someone explain it to me with some commands?

Thank you very much

[Ram]

row count

Do you mean raw counts? Google "raw count to TPM" and you'll find tutorials that you can use.

[dsull]

If you run your FASTQ files through kallisto, you automatically get TPMs and everything you need for CIBERTSORTx. Just read a tutorial on how to use kallisto.

Also, you don't 'need' to use kallisto -- as Ram mentioned, you can approximate TPMs from your raw counts. Again, read a tutorial on this.

[Ram]

Hi michelafrancesconi9

Why did you delete this post?

Answer 1

[swbarnes2]

If you somehow have transcript counts, then you can look up transcript lengths, and calculate everything fine. But since DESeq2 takes gene counts, not transcript counts, I'm guessing you don't have that.

how I can find the gene length

Well, that is the question, or rather, how to calculate the transcript length, or the effective transcript length, given that every gene will be represented by different proportions of different transcripts of different lengths. And none of that is captured with a simple gene count, and while you could look up what the different transcript lengths are for every gene, no online database can tell you what the different proportions of each transcript are for your samples.

Sure you could find some number and just pretend that is the transcript length, but I don't really see the point.

Best to start over, and use something like Kallisto or RSEM that works out the proportions of the transcripts, knows their length, and will output the TPM for you.