Dear Team,

For Differential expression, we are using tweeDEseq method which will take read counts as input and normalize counts using TMM method and perform pair wise comparisons.

I have used Flux-capacitor for quantifying transcripts which gives read counts and RPKM values as output.

here is the few lines of output from flux-capacitor

18 ENSEMBL transcript 158483 213739 . + . transcript_id "ENST00000261601.5"; locus_id "18:158483-213739W"; gene_id "ENSG00000101557.8"; reads 4644.136719; length 4156; RPKM 2497.995850

18 ENSEMBL transcript 158483 213739 . + . transcript_id "ENST00000383589.1"; locus_id "18:158483-213739W"; gene_id "ENSG00000101557.8"; reads 2171.213379; length 4051; RPKM 1198.126099
18 ENSEMBL transcript 158536 211350 . + . transcript_id "ENST00000400266.3"; locus_id "18:158483-213739W"; gene_id "ENSG00000101557.8"; reads 72.650101; length 1681; RPKM 96.611931

And I have two questions with respect to input counts for tweeDEseq package.

1. Can I use RPKM values as input for tweedeseq package?

2. As RPKM values already already normalized counts, can I consider "reads" section as raw counts from flux-capacitor output and Is it ok to use the same as input to tweeDEseq?

Please suggest me.

Thanks In Advance

Fazulur Rehaman

Dear Fazulur,

1. No, you can not use RPKM values as input to the tweeDEseq package. This package implements a statistical model based on Poisson-Tweedie distributions for count values and RPKM values are not count values.

2. No, for the same reason I give on your first question.

As a side note, while Flux Capacitor was one of the first available RNA-seq quantification methods, it seems that its development has been discontinued and other methods, such as Salmon or Kallisto, have been more recently developed for that purpose and are actively maintained. One of the authors of Kallisto described in a blog post why Flux Capacitor performs worse than other methods. Others in this forum may have additional recommendations but since this may be considered out of the realm of Bioconductor, you may be better off asking or googling elsewhere if you have further questions about RNA-seq quantification methods.

cheers,

robert.