MAQ mapping quality scores

[fadista]

Hi,

Just want to be sure about the MAQ mapping qualities.

I have a lot of reads with the mapping quality of 99. Is this the Phred-scaled probability that the read alignment is wrong? And if so, the reads with mapping quality of 99 will have a probability of being wrong equal to p = 10 ^(-90/10) = 1e-09 right?

[ECO]

That is my understanding. I am still confused a bit as there is some discussion in the comments of some parts of maq that there is 10 added to the mapping qualities in some parts.

Maybe lh3 will chime in if he sees this.

[zee]

As far as I understand the mapping quality score is related to:

1) sequence quality of the aligned read
2) number of locations the sequence maps to

Ih3 is best qualified to give a more comprehensive answer.
If your score distribution is skewed to 99 then you either have very high quality data or perhaps there is a bias for a particular region that was sequenced from the sample.
AFAIK MAQ assigns a quality of zero to randomly chosen matches, so I think it's unlikely that your 99 mapped to multiple loci. Do you observe a similar score distribution with other aligners?

[Layla]

This is a very interesting problem. I have run maq match and see that 82% of my reads are correctly paired. Secondly, 40% of my reads have a mapping quality of 0 and thus mean these reads have multiple hits. Thirdly only 2% of my reads have a mapping quality of 30 (error 0.001). However, the mapping quality values range from 30 to 99. What do these represent? Are they considered reliable and hence kept for further analysis?
Any help or advice/discussion would be very useful.

L