RNASeq pipeline

[Sow]

Hi,

I was wanting to know if it is alright to normalize my RNA seq data before mapping the reads to the genome? I ask this because I have the raw file, however the mapping was performed by a bioinformatician who did not provide us the mapped reads file but just the output from after running all the pipelines.

Also, can I run a differential gene expression analysis on the raw reads - even before mapping? Would these results be accepted as valid fold changes if performed on the original files?

Thank you

[Brian Bushnell]

Hi,

I was wanting to know if it is alright to normalize my RNA seq data before mapping the reads to the genome?

No.

I ask this because I have the raw file, however the mapping was performed by a bioinformatician who did not provide us the mapped reads file but just the output from after running all the pipelines.

He should provide you the mapped reads upon request.

Also, can I run a differential gene expression analysis on the raw reads - even before mapping?

No, that's impossible. How can you know which read goes to which gene without mapping? That's the definition of mapping.

[Sow]

Thanks so much Brain,
I should have mentioned that this is an small RNA Seq data. So mapping tells us if the reads come from the genome or if it's a random degraded transcript. However it is right not to run differential expression analysis on the raw reads - isn't that what deseq requires us to do?
Also- I was hoping I could obtain the fold change of the miRNAs ( obtained after running the pipeline) from the differential expression Analysis of the raw reads?

[ddb]

Note that raw reads and raw read counts (what DESeq requires) have a different meaning. Raw reads suggests your data as it came from the sequencer. Raw read counts is the number of reads that map to the genes / features of interest without normalization for factors like library size. You cannot get raw read counts without mapping first.

[WhatsOEver]

No, that's impossible. How can you know which read goes to which gene without mapping? That's the definition of mapping.

Not completely true. K-mer counting based tools like "salmon" or "kallisto" can perform differential gene expression analysis without the "classical" mapping. But you'll need to have well curated transcriptome sequences and identification of novel transcripts or isoforms is obviously also not possible.
Nevertheless, the tools work with "raw reads" as the op requested

[Brian Bushnell]

Not completely true. K-mer counting based tools like "salmon" or "kallisto" can perform differential gene expression analysis without the "classical" mapping.

Or Seal

But I would argue that what those tools do is mapping, just without an alignment phase.