I noticed that the gene expression count from HT-Seq / featureCount followed by STAR is much more higher than what RSEM got. It seems RSEM assign more much alignments as non-primary.
I did some digging and got vague impression that this is because RSEM consider more than just counting the reads aligned to gene region like HT-Seq / featureCount does.
I never got a clear explanation what causes this difference. Can anyone share some comments. Really appreciate.