Why do I see differences in calls between version 3.3 and 4.0?
Archer Analysis 4.0 now employs a de novo assembling strategy for all fusion and oncogenic isoform calling. While this can produce results that vary in statistics from prior versions, it's proven to be more accurate and should result in fewer ambiguous calls. Since this is a parameterized option, it can be toggled on or off in user settings (we recommend on for the reasons listed above).
Some benefits of the de novo assembling strategy are:
- We see more consistency in read clustering. Because de novo alignment depends only on the similarity of reads to each other and not to the genome (in the first step), the alignments surrounding the breakpoint tend to be longer and of higher quality.
- Annotation is more robust because the de novo alignment reduces the occurrence of mis-aligned and semi-repetitive reads to an incorrect genomic location.
- De novo alignment usually leads to more "collapsed" fusion calls. That is, in version 3.3 and version 4.0 you can sometimes get the same fusion breakpoint reported multiple times due to slight differences elsewhere in the reads, but away from the actual breakpoint location. The issue is less prevalent in version 4.0, although it may still occur if a fusion is called by more than one GSP2. This means that you don't have a given number of supporting reads split out over multiple calls of the same fusion, thus the metrics for the collapsed fusion have more reads, more unique start sites and a higher percent GSP2. Ultimately, this leads to higher confidence calls.
For Research Use Only. Not for use in diagnostic procedures.