Don'T Good Base Quality Scores = Good Mapping Scores? (Maybe A Reference Error?)
2
1
Entering edit mode
10.9 years ago
diltsjeri ▴ 470

Hi,

While analyzing some amplicons we decided to increase the mapping stringency in the post sequencing process to ensure the data was free of debris. As we increased the mapq value to 18 (using samtools) we noticed the majority of a particular amplicon's mapq values were lower than 10.

Amp1 below shows normal data from a different amplicon and Amp2 is the the amplicon we are having questions about.

Amp1        Amp2
mapq count  mapq count    
<=10  11   <=10  986
11-20 7        11-20 5
21-30 3        21-30 0
31-40 1        31-40 0
41-50 0        41-50 1
51-60 6        51-60 0
61-70 12    61-70 1
71-80 79    71-80 1
81-90 316    81-90 0
91-100 100    91-100 0

We looked into the trim function and found that the default cutoff is 16 and the window size is 30. This means that all the base accuracy is 97.5%.

Don't good phred values also give a good mapping score? This is assuming my reference is OK and it seems to be.....

Any advice would be appreciated. Thanks.

Three different alignments of the same problem amplicon. The reads are in order from highest to lowest mapq values. Notice the drop off at the 3

mapping quality reference • 4.1k views
ADD COMMENT
6
Entering edit mode
10.9 years ago

Mapping quality is also a function of the underlying reference sequence. Flawless reads aligning to accurate sequence can yield low mapping scores if the underlying sequence is repetitive. The repetitiveness means that you can't actually know if the read is aligned to the correct place; since it maps quite well to the other places that share the same, or nearly the same sequence.

ADD COMMENT
0
Entering edit mode

Great point. I could see that happening with our reference. What's a better way to clean up our data if not mapq?

ADD REPLY
5
Entering edit mode
10.9 years ago

There's no required correspondence between phred score, which relates the likelihood that the base was incorrectly called, and MAPQ, which relates the likelihood that the alignment is correct. If amp2 has generally low mappability, then it won't matter how good your reads are.

Edit: I should have updated before replying, swbarnes2 beat me to it!

ADD COMMENT

Login before adding your answer.

Traffic: 1843 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6