[BioC] Feature request for Rsubread::featureCounts: read length adjustment

Ryan rct at thompsonclan.org
Tue Apr 8 08:58:49 CEST 2014

Hi Wei,

> I'm not entirely sure what you are trying to do. But would extending the genomic regions you use in your summarization achieve the same effect?

No, that would effectively extend both ends of each read symmetrically. 
I want to keep the 5-prime position of the read the same, but change 
the length. So if the effective fragment length was set to 150, then a 
100-bp read mapped in the forward direction at position 500 would 
overlap a peak that starts at 625, but it would not overlap a peak that 
ends at 475.

> For your second request, maybe you can do a filtering after you get the read counts, which is pretty straightforward to do?

I think you've misunderstood what I'm asking here. It's kind of hard to 
explain in words. I mean that currently, if there is even 1 bp of 
overlap between a read and a feature, featureCounts will count it. I'm 
saying that it would be nice to be able to be more stringent by 
requiring more than 1 bp of overlap. E.g. require 50 bp of overlap for 
a 100bp read to count it, or even count only reads that fall completely 
within a feature (i.e. 100% overlap).

Now that I think about it, I could implement the first request and part 
of the second one if I could provide the reads in e.g. a GRanges object 
or a text file that just has columns for chromosome, start, end, and 
strand (or a bed file, etc.). Then I could pre-process my reads to 
adjust the fragment lengths however I want. However, the featureCounts 
help indicates that bam (or sam) is the only acceptable input format. 
Is this correct, or is there another way to provide the input reads?


> On Apr 8, 2014, at 11:19 AM, Ryan C. Thompson wrote:
>> Hello,
>> I would like to request a simple feature for Rsubread's featureCounts function that would make it more useful for ChIP-Seq applications. I want to use featureCounts to count the number of reads falling in each of my called peaks. However, each read represents a DNA fragment of a specific length, which can be estimated by cross-strand correlation analysis or known a priori. In my case, it is the length of one nucleosome, i.e. 147 bp. So I would like to treat each read as being 147 bp long for the purpose of computing overlaps, since the number of bp sequenced is not representative of the fragment length. Would it be possible to add a parameter to featureCounts to allow this adjustment? Also, an additional feature that would be nice to have, but is less important, would be the ability to require that a certain percentage of a read overlaps a feature before counting it.
>> Thanks for listening,
>> -Ryan Thompson
> ______________________________________________________________________
> The information in this email is confidential and inte...{{dropped:4}}

More information about the Bioconductor mailing list