For pooled SNP detection, we only considered reads with a MAQ mapping quality of 20 or more and with three or fewer mismatches to the reference sequence.
These observations call for more accurate assessment of mapping quality for the mappings provided by these tools.
The properties include sample identifier, strand, read group, mapping quality, base call at a particular position, pair distance and orientation, custom tags and more.
Reads with mapping quality zero were discarded.
A recent benchmark test highlighted that filtering using BWA generated mapping qualities with a stringent threshold removed more than 80% of the false alignments at the cost of 1% loss in sensitivity compared with the unique mapping strategy [ 80, 82].
MAQ and BWA generate mapping qualities.
We defined high-confidence mapped tags as those with a mapping quality score more than 20.
The higher the mapping quality score, the more accurate an alignment is.
These results are consistent with the precision/recall curves shown above alignment tools report many incorrect mappings with high mapping quality, and LoQ uM's more conservative quality scores indeed reduce false-positive SNP calls.
LoQ uM can work with a wide range of alignment tools to run the aligner for the user, compile the mappings returned by the aligner, calibrate the quality of these mappings and return the list of mappings with more reliable mapping quality scores.
