I have generated nanopore long-read consensus SV callsets using minda by taking the union of three somatic variant callers (severus, savana, and nanomonsv)...But these callers do not classify variants in the same way.

I don't want to just leave the results as unclassified SV breakends as it would be helpful to see how many of each SV type there within the cohort.

Is there a method/tool or otherwise accepted way to classify merged SVs?

Or should I just take the SVType from one of the callers?