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,
First of all, thanks for developing this amazing tool.
While I was tryin
8681
g to utilize minisv.py on my tumor-normal matched data, the relation between POS, SVLEN, END position values does not match.
From the example result above, relatively short Deletion calls with actual deletion sequence on vcf show well-matched POS, SVLEN, END relationship.
(POS + |SVLEN| - 1 = END)
However, for the another calls of large deletions, the equation above does not match.
As it's not a huge difference, annotations might not differ much which value I choose, but I want to ask which one should I use for my analysis.
I agree that HiFi reads are often not long enough to call those extremely large Indels, and boundaries are always the problem in SV calling, but I wish at least the values on single variant call matches each other. (even it is not the truth)
Hi @Yoonsung1203, thanks for reporting this and for including the example. We reproduced the same pattern with synthetic input using minisv genvcf. For CIGAR-based deletions, the exporter writes POS as the first deleted base. For split-read deletions, POS marks the base before the estimated deletion interval. The split-read SVLEN calculation counts both aligned flanking bases and adjusts for the gap or overlap between the read alignments.
If your VCF was generated with minisv genvcf, use the reported coordinates for annotation. In 0-based, half-open BED coordinates, use [POS - 1, END) for CIGAR-based deletions and [POS, END) for the estimated span of split-read deletions. In the original .msv file, the third column of a deletion record is numeric for CIGAR-based calls and >> for split-read calls.
Could you share the miniSV version and command you used? If you still have the .msv file, one corresponding row with read and sample names removed would help us verify your example.
Activity
qinqian commentedon Nov 16, 2026
Hi @Yoonsung1203, thanks for reporting this and for including the example. We reproduced the same pattern with synthetic input using
minisv genvcf. For CIGAR-based deletions, the exporter writesPOSas the first deleted base. For split-read deletions,POSmarks the base before the estimated deletion interval. The split-readSVLENcalculation counts both aligned flanking bases and adjusts for the gap or overlap between the read alignments.If your VCF was generated with
minisv genvcf, use the reported coordinates for annotation. In 0-based, half-open BED coordinates, use[POS - 1, END)for CIGAR-based deletions and[POS, END)for the estimated span of split-read deletions. In the original.msvfile, the third column of a deletion record is numeric for CIGAR-based calls and>>for split-read calls.Could you share the miniSV version and command you used? If you still have the
.msvfile, one corresponding row with read and sample names removed would help us verify your example.