|
|
Mapping Selection
Following a seed-based search, the Isaac Aligner selects the best mapping among all the candidates. For paired-end data sets, all mappings where only one end is aligned (called orphan mappings) trigger a local search to find additional mapping candidates. These candidates (called shadow mappings) are defined through the expected minimum and maximum insert size. After optional trimming of low quality 3' ends and adapter sequences, the possible mapping positions of each fragment are compared. This step takes into account pair-end information (when available), possible gaps using a banded Smith-Waterman gap aligner, and possible shadows. The selection is based on the Smith-Waterman score and on the log-probability of each mapping.