identifying miRNA in RNAseq

Question

[CHANG]

I have Illumina hiseq pairend RNA sequencing on tumor samples. I am wondering if it is possible to identify microRNA expression in that data.

Answer 1

[Lorena Pantano]

Hi,

The protocol for RNA seq is different. First, the kit to purified RNA could be bias to genes and not miRNA. Second, normally there is size selection before sequencing that will skip miRNAs. Maybe you can detect precursors, but the expression of the precursors it could not to be correlated to the miRNA expression, and don't know how is the abundance of these compare to the rest of the transcriptome, so maybe the values are low.

Answer 2

[Jelena Aleksic]

You might pick some up, but really, an mRNA-seq experiment is designed to pick up protein coding genes. If you are interested in a detailed look at miRNA expression, it is better to do a dedicated sRNA-seq experiment. As Lorena already mentioned, the protocol is different, and different size selection is applied. I would also use a different sequencing length, and the presence of adapters in the sequence suddenly becomes more of an issue than before.

Answer 3

[Dan D]

For starters, I would align your data to a known miRNA reference. I prefer miRBase.

[Lorena Pantano]

here some evaluation for this: http://lorenapantano.wordpress.com/2014/02/28/mirna-annotation-tools-which-is-the-best/, just use the first read of your data.

[Dan D]

Interesting post. I'll definitely be using your tools the next time I need to study miRNA expression in-depth. For this particular question, I'm suggesting aligning directly to miRNA sequences as an answer to the question "is there miRNA in my reads?" More sophisticated analysis would follow a "yes" answer.

[Lorena Pantano]

you are right, the main reason why I would use miraligner is because you get a sense whether the reads map to the mature or precursor directly allowing the detection of isomirs quite easily, just if you want to make sure what is the amount of miRNAs that could be in that data (that I guess low, but not sure)

But yes, mapping to mature sequences with bowtie, bowtie2 or any other fast aligner will give you the answer if there is any match. So, agree with you, of course.