walaj
comparing svaba to snowman
We are reaching out to understand any major differences between svaba and snowman, that could cause the calls to be very different. We recently downloaded some PCAWG results that were generated using snowman, and ran svaba starting from the same bams, and then compared the two results using mergeSVvcf. There are about 40 samples. Consistently across the samples we found that there are many variants that are unique to each caller, and less variants shared by both callers than we would expect. Perhaps the snowman calls we downloaded were filtered, and therefore missing svaba calls, but that does not explain why our unfiltered svaba calls are missing a lot of snowman calls. We are happy to provide some examples if you want to look into it further.
