Mappability (Mapping Quality) Score Of Unique Aligned Reads In Bam Files

Question

[alon.goren]

Hi,

We want to get reads that are mapping to only one loci in the genome. Looking at this http://sourceforge.net/apps/mediawiki/samtools/index.php?title=SAM_FAQ#I_want_to_get_.60unique.27_alignments_from_SAM.2FBAM.

It seems like we can use the mapping quality as a way to filter.

Our question is what would be a mapping quality score that would provide a stringent mapping (i.e. aligned to only one loci)?

Thanks!

Alon.

[Zev.Kronenberg]

take a look @ Extract Only Uniquely Mappable Reads From Bam Files

Answer 1

[lh3]

In theory, any alignment with mapQ>3 can be considered to be unique, but you will still get a lot of wrong alignments for mapQ<10. There is no single threshold that suit every application. Some tools also build mapQ into model and you do not need to filter at all.

If you have to choose one threshold, I would say 10 for a typical aligner. A more scientific way is to choose multiple threshold and to see if your results are sensitive to the threshold. Probably little.