Gene Sums for Chip-Seq Data

[BenS]

Hi all,

I'm an amateur at bioinformatics, so I'm having a bit of trouble with trying to get some information out of Chip-Seq data. I would like to know if there is a way to get the sum of the amount of Chip on genes based on a list (like the tables from UCSC).

I have .bam files and .wig files, and my current solution is that I'm trying to create a JAVA program to sum up the amount based on the .wig file with the reference table, but I was wondering if there was a more efficient / accurate option.

Basically, I'm looking for gene by gene normalized sums similar to the FPKM I get from cufflinks for RNA-seq, but for Chip-seq. (Is feasible to run cufflinks on Chip-Seq data?)

Up until now I have only been able to get any sort of sums out of the wiggle files was by taking the summing the numbers from SitePro's text file with profiles, but that will only get me sums around a certain point, not for the gene bodies.

Is this clear?
Can anybody help me?

I would appreciate any assistance anyone could provide.

-BenS

[markrobinsonca]

Hi Ben,

Do you use R/Bioconductor? If so, we have a couple accessory functions in the Repitools package [1] to go from a set of BAM files to read counts at specified regions. There are a few options -- annotationCounts() if you want to count reads around genomic features (e.g. 1000bp upstream to 500bp downstream of a TSS) or annotationBlocksCounts() for counting reads for specified regions (e.g. gene bodies, as you describe, or tiled regions genome-wide).

I doubt that running cufflinks on ChIP-seq will be fruitful.

Best,
Mark

[1] http://www.bioconductor.org/packages...Repitools.html

[mudshark]

those operations are pretty straight forward using IRanges in bioconductor. Have a look at the findOverlaps function for example.

I would be very careful with any normalization as there is no valid approach. FPKM is particularly problematic as gene-length differ substantially and chromatin features are frequently unevenly distributed along genes.

it is furthermore essential to compare your counts to a control sample (input chromatin) as the number of reads on a gene correlates with the activity state of a gene. in other words, you get more reads from genes that are more active as chromatin fragment release on these genes is increased.

[BenS]

Thanks, I'll try both those suggestions, I'm assuming this means I should try importing the bam files into R. I have used R samtools before, but is that the method I should be using?