Closed sowalsky closed 3 years ago
In our own workflows, we usually use fade as a last step of bam processing before variant calling. For us, this is after duplicate removal, indel realignment, and base quality score recalibration. As far as GATK3 to GATK4 differences, I have not tested fade for use cases where indel realignment is not performed. However in figure 3 of our publication, we show that fade seems to have a negligible effect on the ability to call INDEL variants. I suspect due to the requirement for the algorithm to find a nearby match to the opposite strand. I think it is safe to assume that whether or not INDEL realignment is done, fade should still be able to identify artifacts from real INDELs.
@sowalsky Closing the issue, but please by all means reach out if your \issue is not resolved or something else comes up. Kind regards.
The documentation does not indicate when FADE should be run when following alignment workflows, such as GATK? Immediately after alignment and qname sort, or after recalibration/mark duplicates? Also, how does FADE work when comparing GATK 3 (with local indel realignment) to GATK 4 (no realignment)?