Using multimaping reads or unique reads on featurecounts?

Question

[omargmc.tak]

I'm working with Rat transcriptome (mRNA) using HISAT as aligner and featurecounts (subread) to count reads using BAM files from HISAT. Featurecounts has the possibility to count only unique reads or multi-mapping reads. What is the best practice, taking into account multi-mapping reads or not? When I do the counting process with multi-mapping reads it increases by 1-4% of reads counted in each sample, approximately, comparing if I only count unique reads. So, I'm not sure what is better from the point of view of what option I have to choose, using multi-mapping reads or only unique reads.

[GenoMax]

Best practice is not to count multi-mapping reads. If only 1-4% of your reads are multi-mapping then you could ignore them. If you want to be able to use multi-mapping reads effectively then look at methods such as salmon and kallisto instead. They work with the transcriptome though which you should have in case of rat.