Mapping short reads

[Babak A]

Hi Alex,

I am trying a new smRNA-seq sample prep protocol for sequencing short 15-35nt RNAs. Currently, I am mapping them using STAR's miRNA ENCODE commands. My reads have the following properties:

• More mismatches/indels at the ends (both 5' and 3') than the center of the read
• More errors and Ns at homo-polymers (naturally)
• Occasionally a couple of bases from the adapter cannot be cut and are left in the read

A large fraction of the reads (40%–50%) are not mapped due to "too many mismatches." Soft clipping reduces this to 10%–20%, but it occasionally cuts a large portion of the read just to avoid a gap open penalty.

Do you have any recommendations? Have you modified your smRNA-seq mapping recommendation since ENCODE?

Many thanks,

Babak

[Alexander Dobin]

Hi Babak,

sorry for the belayed reply.

You can allow more mismatches to capture the reads with more errors and unclipped bases.

STAR cannot detect indels in very short reads.

[Babak A]

Thanks, Alex! Sorry for my late reply as well. I was not notified that you have responded.

In general, how do you view soft-clipping for mapping short reads? For example, what happens if we don't trim all (short) adapters? Will it affect mapping quality? In our situation, we have two major issues:

• The 5' adapter removal might remove two-three bases from the 5' end of the read
• The 3' adapter removal might remove one additional base from the 3' end of the read

If we don't aggressively trim the adapters and allow soft-clipping, it seems more reads will be mapped.

Babak

[Alexander Dobin]

Hi Babak,

the danger of keeping adapter bases is that they could limit the mapping to fewer loci, compared to the true mapping to multiple loci after the adapter bases are removed.

If you want to be extremely careful, you can try to map with and without adapter trimming, compare the alignments, and design some rules of how to choose the best alignments.

Cheers

Alex

[Babak A]

Thanks, Alex! This is indeed helpful.

Babak