How do I tell featureCounts to ignore fragments that overlap with more than one feature?

Question

I am using featureCounts from the subread package to count the number of fragments per gene in an RNA-seq experiment. I feed it a BAM file with paired-end reads aligned to a reference genome. In the help page of the executable, which I have installed using conda install subread, there is a section about overlaps, including for example the -O option which will assign the reads/fragments to all overlapping features. What is the default behaviour? Assigning to reads/fragments to the feature with largest overlap? not assigning multi-overlapping reads/fragments?

I would appreciate your help on this.

OS: macOS X

featureCounts version: 1.6.3

Answer 1

If a fragment overlaps multiple genomic features then it's unassignable, which is the appropriate way to handle ambiguous mappings when the resulting counts are then given to tools like DESeq2.