Count reads mapping to non-feature area

[tirohia]

I have a bam file containing arabidopsis short RNA reads that I have mapped to the TAIR10 genome. Is there an easy way to get counts of reads that overlap each other and the corresponding region/sequence that they overlap. Essentially, I'm wanting to group all the reads in each bam file into groups that map to the same region of the genome to get a pseudo count (don't worry, I'm not using this for anything serious), I'm exploring at the moment.

I mapped them to the sRNA from mirBase and essentially nothing maps (0~20 reads). There's a reasonable number of reads mapping to the genome, so I want to find areas where there's significant numbers of those reads clustering.

I have a nagging feeling that this is a silly question with an easy answer, but the only way I can think of doing it at the moment is to align all of the mapped sequences with each other and count how many have decent overlaps. Which seems silly.

Cheers
Ben.

[GenoMax]

http://bedtools.readthedocs.org/en/latest/content/tools/genomecov.html

or featureCounts: http://bioinf.wehi.edu.au/featureCounts/

[Michael.Ante]

Hi Ben,
you can also try htseq count here.
If your transcript/gene annotation has additional information like gene_biotype you can count how many (uniquely mapping) reads map to protein-coding genes etc. All reads which do not overlap with your annotation are counted as "no-feature".
Cheers
Michael

[tirohia]

Thanks Geno - I've managed to get something sort of close to what I'm trying to do out of bedtools - I should have thought of that.

Michael - I'm actually looking for areas with no annotation that have disproportionately large numbers of reads, which, if understand HTSeq correctly, would all just get lumped into the no-feature category.

Ben.

[GenoMax]

Did you create an "anti-bed" file (sort of like Genome - real bed) to get the regions you are interested in?

You could also use some kind of moving window read count and then exclude regions that are coding etc. See this thread for some inspiration: https://www.biostars.org/p/58781/

[Michael.Ante]

Hi Ben,
First, I like GenoMax's idea.
If you have non-annotated clusters attracting a lot of reads, you may like to look at the repeatmasker. This tool gives you regions defined by repetitive sequences; inter alia rRNA, simple repeats, tandem repeats, and many more.
AFAIK, there's a compiled annotation for Arabidopsis out. If you can't find it, you need to download the software and run it on your genome annotation.
Cheers
Michael