Tweet
Hi all,
I am working with Illumina HiSeq paired-end reads mapped with Bowtie2 to a bacterial reference genome in 3 biological replicates. By now, I pursued two strategies to quantify the mapped reads per gene.
(1) Used HTSeq (union- mode) to get read counts followed by TMM normalisation
(2) Used Cufflinks to calculate FPKMs (with and w.o. length correction)
Since I have only biological replicates but no different treatments/conditions my aim is to make more general statements about gene expression (like mean values) under this one condition and not necessarily assessing differential expression between the biol. replicates.
I'm new in the field of RNA-seq but hardly realised that there are lots of tools but few standards. However, consensus seems to be that comparing gene expression within one genome requires normalisation by gene length and comparisons between same genes but different samples normalization by library size.
But as I tested correlation between HTSeq counts and gene length I found no correlation at all.
(A) is it ok to normalise for length despite of a missing count vs. gene_length correlation
(B) What would you recommend for a quantification/ normalisation strategy in my specific case?
I am working with Illumina HiSeq paired-end reads mapped with Bowtie2 to a bacterial reference genome in 3 biological replicates. By now, I pursued two strategies to quantify the mapped reads per gene.
(1) Used HTSeq (union- mode) to get read counts followed by TMM normalisation
(2) Used Cufflinks to calculate FPKMs (with and w.o. length correction)
Since I have only biological replicates but no different treatments/conditions my aim is to make more general statements about gene expression (like mean values) under this one condition and not necessarily assessing differential expression between the biol. replicates.
I'm new in the field of RNA-seq but hardly realised that there are lots of tools but few standards. However, consensus seems to be that comparing gene expression within one genome requires normalisation by gene length and comparisons between same genes but different samples normalization by library size.
But as I tested correlation between HTSeq counts and gene length I found no correlation at all.
(A) is it ok to normalise for length despite of a missing count vs. gene_length correlation
(B) What would you recommend for a quantification/ normalisation strategy in my specific case?