mtRNA expression distribution in RNASeq

[ZoeG]

Hey, Any one experienced in mtRNA (mitochondrial RNA) expression?

I mapped my data to mtRNA and found that the counts of protein-coding RNA are much higher than that of tRNA and rRNA (hundreds or thousands fold).

Is this a normal mitochondrial RNA distribution?
Thanks

[dpryan]

That kind of depends on the species, tissue, and methods used to extract the RNA and make the library, doesn't it?

[sara_ahmed]

Agree with dpryan. It depends on all of those things. Were these prepared after polyA selection? That tends to have the highest mtRNA carry through because of polyA tails on mtRNA. The levels can fluctuate between cell types but its not uncommon to see higher levels of mtRNA compared to rRNA reads.

Sara Ahmed, PhD | Director of Sequencing | Cofactor Genomics
3141 Olive St. | St. Louis, MO 63103 | tel. (314) 531-4647

[ZoeG]

Thanks, dpryan, sara.
The total RNA was extracted using Qiagen RNeasy mini kit from mouse skeletal muscle. The library was prepared following illumina TruSeq sample preparation. I suppose these steps are common for RNA preparation and library preparation.

The way how I mapped to mtRNA is some tricky.
I didn't map my data to mtRNA directly.
Raw data was mapped to a GTF containing chr1~chr19+chrX,Y, but not chrM firstly.
Thus, genes from chrM were left in the unmapped reads. Then, I mapped unmapped reads to a chrM GTF and got the counts of mitochondrial genes.
Mapping chrM genes first time, not sure whether these operations will induce bias..
Comments?

[dpryan]

Skeletal muscle has a lot of mitochondria, so I'm not surprised that you get an enrichment of that. With the Qiagen kits, remember that the columns themselves do some size filtering, so you'll lose much of the RNA <200nt (namely tRNA, microRNAs, and the smaller rRNAs). Those two alone could yield the results that you see.

BTW, I would encourage you to include chrM in your reference during alignment. As a general principle, you increase false-positive alignments if you align reads arising from the entire genome to only a subset of the genome. Granted, chrM is pretty small, so that bias is probably not large, but it's simple to just include that in the reference fasta and GTF/GFF.

[ZoeG]

Thanks, dpryan. I understand that align entire genome reads to a subset of genome will causes false positive outputs, especially mitochondrial genes are relatively small. So I am thinking whether it is okay to align those big genes first and then align unmapped to chrM genes. But you are right, it worths a try to include chrM in the reference.
Another reason for me to do this is that my raw data seems a little weird. If I align the raw data to whole genome (TopHat -G), it generated tons of 'novel' junctions which is probably not true. I have to add some conditions, for example, align to known genes only. I am not sure whether these conditions will exclude chrM genes, will try soon.