bimodal distribution of bam mapping quality

Question

[irritable_phd_syndrome]

I have some RNA-Seq data that I analyzed with tophat2. The command that I used to generate it is

/path/to/tophat-2.1.0/tophat -p 20 -o ouputdir --library-type fr-firststrand /reference/homo_sapiens/GRCh38/ensembl/Sequence/Bowtie2Index/Homo_sapiens.GRC38 my_trimmed_data.fq.gz

This output a file, accepted_hits.bam with 50e6 aligned reads. When I plot histogram of the mapping quality scores, roughly 27e6 reads have a mapping quality value [0-3] and 23e6 reads have a mapping quality value of 50. There are _no_ values in between.

How should I interpret this bimodal distribution of mapping quality scores? This seems very strange to me.

Answer 1

[John]

This is normal :)

Tophat2 Mapping Qualities

Answer 2

[dariober]

Besides Tophat, where intermediate mapq is avoided by design, the mapping quality from other aligners (e.g. bwa) tends to be markedly bimodal. I.e. reads tend to be either very unambiguously mapped somewhere or they can map equally well at multiple places.