the --fraction option of featureCounts

Question

[Apex92]

Hi all,

I have a very specific question for the -fraction option of the featureCounts tool. I am analyzing single-cell data and only focusing on unique mapping reads. What I am aiming is that to weigh the counts for reads if they map to two different annotations that have an overlap.

So for example imagine a read gets mapped to two different annotations that have an overlap, then the total count for that read should be 1 (0.5 each) not 2 - is this what the -fraction option does?

Or is there a better solution for this?

Another thing about the -fraction is that it should be used with the -M (for multi-mapping reads) - while I have only allowed for unique mapping reads in my mapping step using STAR - can this cause a problem?

Thank you for your inputs in advance.

Answer 1

[Carlo Yague]

What version of featureCounts are you using ? In version 2.0 --fraction can be used with either the -O or -M option (or both). With the -O option, -M is not needed, and if a read overlaps two feature, 0.5 will be assigned to each of them. Here is the relevant documentation:

  --fraction          Assign fractional counts to features. This option must
                      be used together with '-M' or '-O' or both. When '-M' is
                      specified, each reported alignment from a multi-mapping
                      read (identified via 'NH' tag) will carry a fractional
                      count of 1/x, instead of 1 (one), where x is the total
                      number of alignments reported for the same read. When '-O'
                      is specified, each overlapping feature will receive a
                      fractional count of 1/y, where y is the total number of
                      features overlapping with the read. When both '-M' and
                      '-O' are specified, each alignment will carry a fractional
                      count of 1/(x*y).

[Apex92]

Dear Carlo thank you for your input - I do use featureCounts from the subread/2.0.0 package. So based on what you have mentioned, then -O seems what I want. But the question is -O also weighs the counts for overlapping features if only unique mapping reads are considered (I excluded all the multi-mapping reads and working only with unique mapping reads)?

[Carlo Yague]

Yes yes, it is not a problem. A read can be uniquely mapped (one genomic position) but overlap multiple features.

[Apex92]

great help - thank you so much :)