Tool for RIP-seq analysis

[honey]

I have a RIP seq data to study the binding of RNA protein. I googled and found only 4 publicaations and searching this forum I found two posts. My question is - Is there a tool or work flow out there for analyzing Illumina Hi seq RNAseq data for specifically understanding the RNA binding proteins.
Alternatively, I am planning to run my regular RNA-seq workflow of running-TopHat> Cufflink and then work with RPKM/ FPKMs
Any suggestion please?

Thanks

[Dario1984]

It's different to RNA-seq. You only get part of the RNA that it bound by the protein. There is no reason to use Cufflinks. I suggest TopHat followed by a ChIP-seq peak calling algorithm. There was a group that planned to publish a RIP-seq pipeline late last year, but I have not seen it released yet.

[honey]

RIP-seq

So you mean after running TopHat I should run MACS or another ChIP- peak caller but how that will be useful? I apologize if someone can through some more explanation that will be great.

So far from this forum ChIP-seq is not good for RIP -seq

http://seqanswers.com/forums/showthread.php?t=17615&highlight=RIP-seq

Used RNA -seq work flow RIPseq

http://seqanswers.com/forums/showthread.php?t=9885&highlight=RIP-seq

This paper used RPKM:

http://www.sciencedirect.com/science/article/pii/S1097276510009676

However it is not clear what bioinfirmatics is done here:

http://www.ncbi.nlm.nih.gov/pubmed/21858232

Any further insight will be greatly helpful.

Thanks

[kopi-o]

You only get part of the RNA that it bound by the protein.

From the (limited) understanding I have of RIP-seq and CLIP-HITS/CLIP-seq, this only applies to the latter. For RIP-seq, as far as I know, you will get longer, less specific RNA fragments. I think using Cufflinks or some other FPKM quantification method actually makes sense, as part of the workflow; some more localized ChIP-seq-type enrichment analysis should probably be performed as well.

[honey]

RIP-seq tools

Thanks kopi-o,
Of the 4-5 RIP-seq publications the follwoing paper:

http://www.ncbi.nlm.nih.gov/pubmed/21051541

Has just aligned with Bowtie and used filteration for reads mapping to exon and introns to coclude the binding patners. I dont understand how you may conclude that? I would love to see response from senior memebers about how should I procede with RIP seq analysis. I apologize for coming back with so many posts, But I want to initate a good discussion on this topic.

[honey]

RIP-seq analysis

I believe I am on wrong track, I dont think TopHat will be ideal as it will be more suitable for splice junctions not for enrichment of specific targets in IP. So I am working with mouse cells and have input, control with isogenic control ab and IP sample with test ab I believe here is my proposed strategy:
Align with Bowtie/ BWA (mouse genome)
Calculate RPKM values per gene and transcript
Remove gene/transcripts which are enriched in isogenic ab.
Calculate the differential expression over input (enrichment in test Ab after removing enriched genes/ transcripts with isogenic ab)

Any suggestion or feed back? Should I be using BWA or Bowtie or adopt some other strategy.
Little lost

[Dario1984]

But your RNA binding protein could be binding to locations on the RNA where a splice junction was and the intron has been spliced out already. You will lose good reads. Could you map to a splice junction RNA reference with Bowtie to see how often this happens ?

[honey]

RIP-seq tool

Agreed Dario that I will lose some information, But I think the main aim is to find in general the binding partners of my protein of interest not that just ones which are specific for junctions. However I am open to suggestions and maybe I am thinking it wrong. I believe using gap junction aligners will be more specific for identification of splicing. TopHat throws out reads which are not specific for junctions. Am I correct?

[kopi-o]

TopHat throws out reads which are not specific for junctions. Am I correct?

No, TopHat gives you both regular genomic alignments and splice-junction alignments.

[honey]

RIP-seq tool

kopi-o Thanks. My bad.

In one of you previous message you mentioned that in this situation I may be using ChIp-seq peak calling tool in addition to RPKM based RNA-seq (differential enrichment) work flow. Would you please comment little further how best you think I can integrate ChIP-se peak caller.
I guess Top hat> Cufflink > RPKM differential enrichment after filtering noise from control Ab enriched regions. Do You think if in parallel I run ChIP-seq peak caller and then overlay those results with above RPKM based differential enrichment?
I understand lot of labs may not be interested in this approach. But any input will be helpful.
Thanks

[kopi-o]

I am not sure about how to approach it, that's why I was a bit vague. I am thinking that transcript-based differential enrichment as you describe can give some information; then there may be other transcripts that don't look differentially expressed overall but where there is some localized over-representation which may be caught by a peak-calling type approach. E g specific overrepresentation of RIP-seq tags in the 3' UTR. Perhaps after defining peaks with e g MACS, a DE-seq type program can be used to assess differential enrichment of those defined peak regions between IP and controls.
Please keep posting your experiences here ;-)

[duffman]

I was working on a similar problem several months ago. My understanding is that RIPseq pulls down entire transcripts, and there is no RNAase that eats away RNA sequence that is not directly bound. Peak-calling seems to be irrelevant in this setting---you get information regarding RBP->transcript association but not binding site locations---you should just get enhanced transcript expression (e.g., as measured by Cuffdiff) for RBP target transcripts.

[Noelya]

Hi everybody,

Since the last post is a little old, I try to get an uptade of what you learned theses last years.

I have done a RIP-seq experiment and I'm looking for a tool (if possible existing on galaxy) which would allow me to quantify reads on the whole genome, not specificaly in the regions definied in the annotation (gff).
I used the tool "Create a BedGraph of genome coverage" followed by "Merge BedGraph files" but I could only see the results for de 10,000 first bases.

So, how do you do these kind of analysis ?

Thank you for your help