Differential expression analysis of transcriptome data

Question

[vijay]

Dear Folks,

I am currently working on brain transcriptome data of a male and female fish obtained using RNASeq. I wish to compare them and obtain the differentially expressed genes. All that I have got is a table containing the Transcripts per milllion (TPM) and the FPKM value for both the samples. I went through a few references which suggested to take log(fpkm) as the differential expression measure. However, I am interested in determining the statistical significance and determine the p-value and FDR of the differentially expressed genes. I learnt that few R packages like DESeq and EdgeR can do this job for me. Unfortunately, I am stuck up in using them as I understand that the TPM or FPKM values are not usable in those packages. Please can someone help me in sorting out this issue and let me know what is required to be done?

I am very much novice in this field and this is my first hands on with RNAseq data. Any help would be greatly appreciated.

Thanks.

[vm.higareda]

Hello Chennai

Did you resolve your problem?. I have a similar doubt. I have the list of fpkm of genes and I want use R packages

Answer 1

[Thibault D.]

Hi,

Packages such as DESeq2, uses their own normalisation methods. You should give them raw counts (the original and non-normalizes ones). You can run tools such as htseq-count or feature-count on your bam files to get your raw counts.

Have a look at the section 1.3.1 of the DESeq2 manual to understand :)

[vijay]

Thanks for your response. Please can you let me know if explaining differential expression only based on log(FPKM) values will make sense. In other words, is that technically correct and acceptable for publication?

[Carlo Yague]

Hi!

It isn't. Shortly, FPKM is a within-sample normalized expression metric and shouldn't be compared between samples.