How to round the "expected counts" to integers?

Question

[joanna]

How can I round the expected counts to use them in Deseq2. The spreadsheet includes Gene names and expected counts and is basically ready to use in Deseq2. I read about Tximport but I was not able to adopt it for my purpose. Thanks for all ideas.

[Rashedul Islam]

If you mean to round floating numbers, you can use round() in R.

[ATpoint]

How were these counts generated?

[joanna]

The 'expected counts' were derived via the RSEM algorithm and I would like to use them as input to DESeq2, but I need to round counts to integers, first. Is there any professional way to do it? Thanks for response.

Answer 1

[ATpoint]

I suggest you use tximport https://bioconductor.org/packages/release/bioc/vignettes/tximport/inst/doc/tximport.html#rsem to read the RSEM quantifications and prepare them for DESeq2. Details in the linked vignette. tximport is from the same group as DESeq2 so integration is seamless. Example code in the vignette. This is superior to any homemade method.

[joanna]

Thanks for response! I've gone through the recommended protocols but still don't know how to implement it. The file I have includes only "Gene names" and "expected counts" for each samples and it's basically ready for use in Deseq2 besides that the counts are not integers. Would you have any suggestion for me?

[simplitia]

Hi others might have other opinion but I think you can just simply round the numbers to nearest integer, this should be good enough. tximport power is to utilize the bootstrapping form things like kalliso but you don't have it here thus I don't think it be that much more useful but I could be wrong.

[ATpoint]

Actually the main point of tximport is to create per-gene offsets that take the average gene length into account in order to correct for bias that comes from expression of isoforms of different length. The bootstraping-thing strongly depends on the upstream method that was used to create the counts/abundance estimates.