Re: [Useq-users] RNA-IP analysis with Useq

[David N. <Dav...@hc...>]

Yes, USeq should work for you.  Many folks are using it for RIP-Seq analysis.  It makes a direct comparison between the number of reads in a given window between a treatment and control.  The total number of reads in each dataset are used to set the expectation of success for the binomial p-value test.  Thus if you have 10x the amount of data in the input then an "even" distribution of reads between treatment and control would be 1:10.  If you have replicas use the MultipleReplicaScanSeqs which wraps the DESeq package to control for overdispersion in your data.

-cheers, D

From: mali salmon <sha...@gm...<mailto:sha...@gm...>>
Date: Thu, 16 Feb 2012 01:51:20 -0700
To: David Nix <dav...@hc...<mailto:dav...@hc...>>, "use...@li...<mailto:use...@li...>" <use...@li...<mailto:use...@li...>>
Subject: RNA-IP analysis with Useq

Dear Useq-users
Is it possible to analyse RNA-IP data using Useq?
I have IP and Input data from RNA that was fragmented to 100nt, and then treated with an antibody against a specific RNA methylation.
I aligned reads to the genome (I don't care of junctions reads, and only 10% of the data couldn't be mapped to the genome), and I would like
to find enriched regions in IP compared to Input.
The idea is similar to ChIP-seq, but I'm wondering if there are special issues I should taking care of since the reads are coming from the transcriptome and not the genome.
For example, in ChIP-seq finding tools like MACS and others, I should give as parameter the effective genome size.
When I run MACS (sorry I haven't tried Useq with this data yet) with gsize=transcriptome size, I get ~2000 peaks, however, when I run it with gsize=genome I get ~22000 peaks. In a genome browser, those peaks that were missed with smaller gsize look real by looking them in a genome browser and comparing them with their Input. Hence I thought of finding a tool that can compare IP directly to Input without taking into account a random expectations calculated based on the genome size. Is it make sense? Is it possible to do it with Useq?
Looking forward to your reply
Thanks
Mali