I have sequencing data from extracellular vescicle done with Qiaseq miRNA library kit with UMI.
I've align the data to the human genome with bowtie2 and removed duplicates with umitools.
Now, I'm trying to identify the miRNA in the data using subreads featureCounts/mirdeep2 and I'm trying also with bedcov. I have different results, bedcov used as it is gives me the highest number of miRNA covered.
Do you think it could be a good solution to identify known miRNA?
Hard to tell without knowing the actual differences between the tools and seeing the outputs, but my 2 cents are that you get leads from subread (I guess featureCounts?) since it doesn't consider multimappers. Also just a recommendation, since you are not looking for novel microRNAs and the human is quite extensively studied, I would recommend aligning to miRBase directly
bedcov from BEDTools can indeed work well for quantifying known miRNAs if you're using it to compute coverage over mature miRNA coordinates from miRBase, especially since it handles BAM files directly and avoids some of the stringency in miRDeep2 that might filter out edge cases in EV-derived data. The higher counts you're seeing likely stem from its less restrictive multimapping allowance compared to featureCounts, which is tuned for gene-level counts and may undercount short miRNAs. I'd recommend cross-validating a subset of your top hits against miRDeep2's known miRNA output to confirm consistency, and consider normalizing for library size and UMI collapse depth to avoid inflation. If your goal is just profiling known miRNAs, this combo should suffice without overcomplicating things.
Hard to tell without knowing the actual differences between the tools and seeing the outputs, but my 2 cents are that you get leads from subread (I guess featureCounts?) since it doesn't consider multimappers. Also just a recommendation, since you are not looking for novel microRNAs and the human is quite extensively studied, I would recommend aligning to miRBase directly