Fusion Filtering

In assays, there are hundreds to thousands of fusion candidates in a single sample. Most fusion candidates (~99%) are false positives. The fusion filtering tool, DNA Fusion Filter (DNAFF), distinguishes true fusion calls from the false positives. DNAFF performs the following functions:

•

Removes spurious fusions including fusions with only one supporting read pair and fusions that overlap with repeat regions, which are more likely to have sequencing errors.

•

Filters nonconfident supporting reads for all fusion candidates based on the following criteria:

–

Filter reads with low-sequence identity with the fusion contig.

–

Filter erroneous reads, which are reads that do not support the fusion. For example, reads that have suspicious supplementary alignment.

–

Deduplicate reads based on UMI information.

•

Applies the following rules for the final fusion output

–

If the fusion gene pair has been reported in COSMIC, it must have ≥ 2 supporting reads.

–

If the fusion gene pair has not been reported in COSMIC, it must have ≥ 3 supporting reads.

–

At least one fusion breakpoint must fall within the 23 target genes.