Maximal Possible Expression Value Of A Transcript

Question

[Steffi]

When having a sample to be sequenced, is there a natural limit for the maximal expression for a specific transcript? So, I mean, can any transcript be expressed millions of times? Or is there a biological limit, which tells me that I cannot expect more than e.g. 1 million counts (assuming one count corresponds to one transcript)?

I have to admit, I am not sure if this question is even well-posed. I would be happy to take any comments!

best,

Answer 1

[Damian Kao]

I guess the theoretical biological limit would be having your entire library consisting of only one transcript. Practically, I don't think any maximum expression limit can be imposed. Are you trying to look for outliers? A better way might be to look at percentage of all read sequenced.

If you are seeing something expressed extremely high, you should look back on how the library was prepared. Was the library ribo-depleted? Are the highly expressed transcripts ribosomal RNA? I've analyzed non-ribo-depleted libraries before where rRNA consisted over 50% of all the reads.

I think this is a case where you can't really use a purely statistical method to identify outliers. You'll have to consider the biology along with the numbers.

You might also consider normalizing your tag counts to length of the transcript for a RPKM value. The idea is that a longer transcripts would produce more fragments in the library, giving artificially more sequenced reads. However, for differential expression where proportion of reads is taken into account, I would stick with tag counts.

[Steffi]

I do not refer to any real data, I do not aim to identify outliers.

It was really just theoretical consideration. I was just wondering if something is already known ...