Bug in featureCounts?

[brinda]

We are using featureCounts for both and single-end and pair-end data. In the command we always have the '-p' flag on, assuming it will be ignored for single-end data. However, we notice that the function of '-M' is affected because of '-p' flag.

When we run single-end data without the -M flag, all the reads with NH > 1 are ignored, as expected. In the test data attached, we get a read count of 4. With -M flag, we get a count of 16, including reads with NH > 1, again, as expected.

However, when we run single-end data without the -M flag but with '-p' flag, the multi-mapping reads (NH > 1) are no longer ignored and we get a read count of 16.

The '-p' should have no effect on filtering multi-mapping reads. I'm attaching a sample sam and gtf file where this behavior may be replicated.

[shi]

As stated in the program usage, the '-p' option should only be used for the summarization of paired-end reads. So you should not use it for the single-end reads.

featureCounts tries to find the pair for each read when '-p' is specified. This causes a problem when single-end reads are provided and '-p' option was incorrectly specified.

In the next release (late Februray hopefully), featureCounts will automatically detect the read type (single-end or paired-end) and it will ignore '-p' option when single-end reads are provided.

Wei