[Bioc-sig-seq] 2D plots

[gbucci]

Hi Michael and Bogdan.
Excuse me for this little off topic.
I'm wondering if some one of you has already done, or have an idea, on  
what should expected to be the correlation between two ChIP-seq  
replicated runs.
I mean: if protein1 and protein2 are actually the same protein and we  
use the windowing function to count how many reads per 1 kb unit we  
have in the two runs I would expect a straight and narrow scatterplot.
The inclination of the fitting line should reflect the abundance of  
reads in each run (the line will move toward the axis were we have  
more reads).
But what function is better suited to calculate such correlation,  
Pearson , Spearman ...?

In a replicated ChIP-Seq experiment, for which I had also a control  
lane (i.e. both IP and INPUT), with a big 1Mb window (mm9 genome), I  
had a Spearman correlation of 0.99 for IP/IP and 0.94/0.96 for the IP/ 
INPUT correlation.
The differences in correlation are smaller than what we expect,  
especially when looking to scatterplots were the differences are clear  
( straight and narrow vs curved and wide).
Changing the window (600K) doesn't change the results, but I never  
tried 1 kb.

The questions are what statistical test is more appropriated and what  
would you expect for a replicated ChIP-seq experiment?
Any Idea?
Thanks everybody.
Cheers

Gabriel


On Aug 26, 2009, at 7:27 AM, Michael Lawrence wrote:

> Well, it would help if we had a few more details, but to calculate the
> number of reads within each 1 kb window of the genome, it's  
> something like:
>
> library(BSgenome.Hsapiens.UCSC.hg18) # assuming human
> widths <- rep(1000, seqlengths(Hsapiens)[1]/1000)
> # assuming 'cov' is a coverage Rle for chr1
> windows <- successiveViews(cov, widths)
> x <- viewSums(windows)/1000
>
> That should get you the numbers you need for that scatterplot.
>
> Michael
>
> On Tue, Aug 25, 2009 at 7:11 PM, Bogdan Tanasa <tanasa at gmail.com>  
> wrote:
>
>> Hi everyone,
>>
>> please could you let me know if there is any way to represent in a  
>> 2D plot
>> the number of ChIP-seq tags / 1kb genome window for 2 experiments ?  
>> eg :
>>
>> on X axis : the number of tags / 1kb window  for protein 1;
>> on Y axis : the number of tags / 1kb window for protein 2;
>>
>> thanks very much.
>>
>> -- bogdan
>>
>>       [[alternative HTML version deleted]]
>>
>> _______________________________________________
>> Bioc-sig-sequencing mailing list
>> Bioc-sig-sequencing at r-project.org
>> https://stat.ethz.ch/mailman/listinfo/bioc-sig-sequencing
>>
>
> 	[[alternative HTML version deleted]]
>
> _______________________________________________
> Bioc-sig-sequencing mailing list
> Bioc-sig-sequencing at r-project.org
> https://stat.ethz.ch/mailman/listinfo/bioc-sig-sequencing

Gabriele Bucci, PhD
Microarrays and NGS Bioinformatics
Consortium for Genomic Technologies
IFOM-IEO Campus
via Adamello 16, 20139 Milan
Italy


gabriele.bucci at ifom-ieo-campus.it
service-arrays at ifom-ieo-campus.it
service-ings at ifom-ieo-campus.it

Int 	45134
Tel 	+39 0294375134