Local realignment

Regions where the readmapping is likely to be improved through local realignment are identified and realigned. These are generally regions where reads do not align perfectly, and the imperfect read alignments are unlikely to be caused by sequencing errors. This is for example the case where long unaligned ends (potentially representing long insertions and deletions) are present in the readmapping relative to the reference.

During local realignment, the following steps are performed for each identified region:

  1. A path graph of k-mers is built. The graph contains paths corresponding to all reads as well as the reference.
  2. The graph undergoes refinement where paths that are unlikely to contain variants relative to the reference path are removed, and additional variants may be added.
  3. Any read that intersects the region of interest is realigned against the alignment graph.

After completion of local realignment, an additional repair step of unaligned ends is performed when running the tools LightSpeed Fastq to Somatic Variants (LightSpeed Fastq to Somatic Variants) and LightSpeed Fastq to Somatic Variants Tumor Normal (LightSpeed Fastq to Somatic Variants Tumor Normal). This is performed only on read pairs that are specifically mapped and serves to align unaligned ends that can be aligned if one mismatch is accepted. The mismatch is, however, not accepted on the last position of the read. Note that this process is not carried out in the first and last 10 bases of a chromosome.