The major limitation of the local realignment algorithm is the necessity of at least one read being mapped correctly according to an indel present in the data. Insufficient alignment data results in suboptimal realignments or no realignments at all. As a work-around, local realignment can be guided by supplying a track of variants that enable the algorithm to determine improvements. Further guidance can be achieved by increasing the amount of alignment information and thereby increasing the chance to observe at least one read mapped correctly.
Reads are ignored, but retained in outputs, if:
- Lengths longer than 50,000 base pairs.
- Crossing the boundaries of circular chromosomes.
Guiding variants are ignored, if:
- They are of type "Replacement".
- They are longer than 100 bp.
- If they are inter-chromosomal structural variations.
- If they contain ambiguous nucleotides.