idr for RNA-Sequencing data

[Sumit Paliwal]

Hi,
I have done RNA-Sequencing with the aim of differential expression analysis. For each of the two groups, I had at least three replicate samples. However, after alignment of the reads I find a huge difference in alignment percentage (ranging from 79% to 30% ) within replicates. This is likely to affect the downstream analysis. I wanted to know if I could use IDR to see correlation between replicates for this data and if yes, how? This will help  me in further analysis.
Thanks.

[Georgi Marinov]

Hi,

Do these happen to be total RNA-seq samples or are they regular polyA+ RNA-seq? There are many possible reasons why a lot of reads do not align. However, that should not really affect DE analysis, because the normalization carried out is only done on the reads that do align. So if half of the reads in one sample do not align because they are of poor quality or you have some mixture of species, or something else of the sort, that does not really matter.

It also generally does not matter if you have a lot of intronic reads if you are using a HTSeq + DEseq pipeline, because DESeq only sees the exonic reads after the HTSeq step.

I would, however, check the overall quality of the samples because there might be major difference that do affect DE analysis, for example, a very strong 3' coverage bias in one replicate in one sample but not in others.

Georgi

--
You received this message because you are subscribed to the Google Groups "idr-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to idr-discuss...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Virus-free. www.avast.com

[Sumit Paliwal]

Dear Georgi,
Thanks.
The RNA-Seq data I have is for Rat tissue polyA+ RNA. I have used several pipelines for analysis that are as follows:

1. Tophat-htseq-edgeR
2. Tophat-htseq-DESeq2
3. STAR-htseq-edgeR
4. STAR-htseq-DESeq2

My concern is that if I use the two replicates with huge difference in alignment percentage, how reliable are my differential expression analysis results? This is because I observe no or very few gene differentially expressed (FDR <0.1 or Padj <0.1).

Thanks for highlighting the fact that I need to look for 3' coverage bias.

Sumit

[Anshul Kundaje]

IDR is not well suited for RNASEQ because RNA expression violates some of the key assumptions. 

Anshul 

--

You received this message because you are subscribed to the Google Groups "idr-discuss" group.

To unsubscribe from this group and stop receiving emails from it, send an email to idr-discuss+unsubscribe@googlegroups.com.

[Ivan Molineris]

Dear Anshul,

can you please elaborate a bit about the assumption that RNA expression violates?

I'm not an expert but if I understood well idr mostly work on ranks, thus should be little sensible to the underling distribution of data.

Thanks

To unsubscribe from this group and stop receiving emails from it, send an email to idr-discuss...@googlegroups.com.