You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi, I'm curious if VarDict generates a file that documents its local realignment processes. Specifically, in instances where VarDict identifies longer indels (exceeding 50 base pairs), there have been observed cases where the supporting reads for these longer insertions do not encompass the entire predicted variant. I'm inquiring if there are any available methods or options within VarDict to generate a file that records these local realignments. Such a file would be instrumental in troubleshooting situations where VarDict predicts extended variants that are not entirely represented by the sequences found in the supporting reads, which VarDict relies upon to make these calls.
The text was updated successfully, but these errors were encountered:
Hi, I'm curious if VarDict generates a file that documents its local realignment processes. Specifically, in instances where VarDict identifies longer indels (exceeding 50 base pairs), there have been observed cases where the supporting reads for these longer insertions do not encompass the entire predicted variant. I'm inquiring if there are any available methods or options within VarDict to generate a file that records these local realignments. Such a file would be instrumental in troubleshooting situations where VarDict predicts extended variants that are not entirely represented by the sequences found in the supporting reads, which VarDict relies upon to make these calls.
The text was updated successfully, but these errors were encountered: