MAPping Quality v.s. Alignment Score (AS) in SAM format

[xiangwulu]

MAPQ: MAPping Quality. It equals −10 log10 Pr{mapping position is wrong}, rounded to the nearest
integer. A value 255 indicates that the mapping quality is not available.

AS i Alignment score generated by aligner


What are the differences between these two? I am confused about this. If I want to filter the alignments and keep the 'good' alignment, which value should I look into?

Thanks.

[dpryan]

Alignment scores are internal to each aligner. MAPQ scores produced by the aligners typically involves the alignment score and other information. Filter by MAPQ.

[xiangwulu]

Thank you.

[dariober]

Just to expand a bit on Devon's answer... Alignment score is a metric that tells you how similar the read is to the reference. AS increases with the number of matches and decreases with the number of mismatches and gaps (rewards and penalties for matches and mismatches depend on the scoring matrix you use). MAPQ is a metric that tells you how confident you can be that the read comes from the reported position.

You can have high AS and low MAPQ if the read aligns perfectly at multiple positions, and you can have low AS and high MAPQ if the read aligns with mismatches but still the reported position is still much more probable than any other.

As Devon says, you probably want to filter for MAPQ, but "good" alignment may refer to AS if what you care is similarity between read and reference.

(Correct me if I'm wrong)

[xiangwulu]

Just to expand a bit on Devon's answer... Alignment score is a metric that tells you how similar the read is to the reference. AS increases with the number of matches and decreases with the number of mismatches and gaps (rewards and penalties for matches and mismatches depend on the scoring matrix you use). MAPQ is a metric that tells you how confident you can be that the read comes from the reported position.

You can have high AS and low MAPQ if the read aligns perfectly at multiple positions, and you can have low AS and high MAPQ if the read aligns with mismatches but still the reported position is still much more probable than any other.

As Devon says, you probably want to filter for MAPQ, but "good" alignment may refer to AS if what you care is similarity between read and reference.

(Correct me if I'm wrong)

Thank you for the explanation.