How To Deal With Multiple Hits When Performing Mirna Expression Analysis

Question

[Doctoroots]

Hi all, ive encountered a problem while working on miRNA expression analysis from an illumina run. usually when i work on the illumina sequence data, i remove all the reads that do not map to anywhere in the reference sequence or that have multiple alignments (non unique). this leaves me with only the uniquely mapped reads to analyze, which reduces the ambiguity of the results and makes the analysis more conclusive.

BUT, it seems that when aligning to miRNA reference (miRBase) there are miRNAs in the reference that are very similar, so similar that most of the reads that align to one of them, aligns to the others as well thus causing the reads to be non unique and to be discarded.

this of course cause a major deviation in the expression results and must be delt with.

how do you suggest i should deal with this issue?

[Michael Kuhn]

"multiple alignment" usually means aligning multiple homologous sequences against each other. I think you mean "multiple hits".

[Doctoroots]

you are right, thank you

Answer 1

[Alastair Kerr]

I normalise by the number of times a read maps, so that a reads only counts once in total. See my explanation in this thread Note that this approach works best if you have told the mapper to show all repeats. Also some tools such as bowtie will do a random assignment for you to get a similar result.

Answer 2

[Michael Kuhn]

I would use same procedure of aligning reads against your reference to align the reference against itself. So if you have A and A' in the reference and both are similar (according to your cutoffs), then if a read hits both A and A' I would not discard it as non-unique.

[Aaron Statham]

Wouldn't a read hitting both A and A' mean that by definition A and A' are similar?

[Doctoroots]

yeah, but thats the issue were discussing, similar elements in the reference cause reads with multiple optimal hits. Michael is suggesting to keep the reads that align to similar elements (that you found by aligning the reference to itself).