% rRNAs discrepancy between sortMeRNA and read count matrices

Question

[jkim]

Hello,

I was working on bulkRNAseq data and ran sortMeRNA on samples. I noticed that around 50% reads of the total reads hit siliva eukaryotic rRNA databases. After that, I obtained a gene count matrix from the STAR --quantMode GeneCounts option and used biomart to pull gene annotation info and merged them.

Here is something I'm a little bit confused about. When I calculated % rRNA from the gene count matrix, I noticed that the % rRNA was about 5% of total uniquely mapped reads. I pull the # reads mapped to the rRNA genes by filtering gene_biotype == "rRNA" | gene_biotype == "Mt_rRNAs" from the annotated gene count matrix.

I was wondering why I observed such a big difference. Any insights or helps would be appreciated.