bcftools norm: Why did it "normalize" this deletion by shifting REF 5' by 29 bp?

[toddajohnson]

Tool: bcftools norm (ver 1.21)
Problem: 276 bp REF sequence for a 275 bp deletion was shifted 29 bp to the left, so normalized DEL does not represent original record. Blat of both the original and "normalized" REF match 100% to their respective locations.

Original record
chr9 125780340 . GGACGGGGCGGCTGGCTGGGCGGGGAGCTGACCCCACCTCCCTCCCAGACGGGGCGGCTGGCCGGGCGGGGGGCTGACCCCCCCCCCACCTCCCTCCCAGACGGGGCGGCTGGCCGGGCGGGGGGCTGACCCCCCCCACCTCCCTCCCGGACGGGGCGTCTGGCCGGGTGGGGGGCTAACTCCCCCACCTCCCTCCTGGACGGGGTGGCTGGCGAGGCAGAGGGGCTCCTCACTTCCCAGTAGGGGAGGCTGGGCAGAGGCGCCCCTCACCTCCCG G . PASS GT:AD:AF:DP:F1R2:F2R1 0/0:17,1:0.055:18:11,1:6,0 0/1:28,12:0.193:40:15,5:13,7

bcftools norm -m-any -f ${HOME}/reference/GATK/resources_broad_hg38_v0_Homo_sapiens_assembly38.fasta
chr9 125780311 . CGGCTGGGCAGAGGCGCCCCTCACCTCCCGGACGGGGCGGCTGGCTGGGCGGGGAGCTGACCCCACCTCCCTCCCAGACGGGGCGGCTGGCCGGGCGGGGGGCTGACCCCCCCCCCACCTCCCTCCCAGACGGGGCGGCTGGCCGGGCGGGGGGCTGACCCCCCCCACCTCCCTCCCGGACGGGGCGTCTGGCCGGGTGGGGGGCTAACTCCCCCACCTCCCTCCTGGACGGGGTGGCTGGCGAGGCAGAGGGGCTCCTCACTTCCCAGTAGGGGA C . PASS GT:AD:AF:DP:F1R2:F2R1 0/0:17,1:0.055:18:11,1:6,0 0/1:28,12:0.193:40:15,5:13,7

[davmlaw]

If you remove the 1st base common between ref/alt - the bases removed in the original variant are

GACGGGGCGGCTGGCTGGGCGGGGAGCTGACCCCACCTCCCTCCCAGACGGGGCGGCTGGCCGGGCGGGGGGCTGACCCCCCCCCCACCTCCCTCCCAGACGGGGCGGCTGGCCGGGCGGGGGGCTGACCCCCCCCACCTCCCTCCCGGACGGGGCGTCTGGCCGGGTGGGGGGCTAACTCCCCCACCTCCCTCCTGGACGGGGTGGCTGGCGAGGCAGAGGGGCTCCTCACTTCCCAGTAGGGGAGGCTGGGCAGAGGCGCCCCTCACCTCCCG

Which if you break it into parts are:

A = "GACGGGGCGGCTGGCTGGGCGGGGAGCTGACCCCACCTCCCTCCCAGACGGGGCGGCTGGCCGGGCGGGGGGCTGACCCCCCCCCCACCTCCCTCCCAGACGGGGCGGCTGGCCGGGCGGGGGGCTGACCCCCCCCACCTCCCTCCCGGACGGGGCGTCTGGCCGGGTGGGGGGCTAACTCCCCCACCTCCCTCCTGGACGGGGTGGCTGGCGAGGCAGAGGGGCTCCTCACTTCCCAGTAGGGGA"
B = "GGCTGGGCAGAGGCGCCCCTCACCTCCCG"

There is another copy of "B" upstream, so the variant represents a change of BAB>B (your variant gets there by describing the deletion of the final AB to be left with B)

The second one also describes a deletion of BAB>B but does so by deleting the 1st BA to be left with B - the equivalent deletion, but left-aligned

Normalized representation of sequence (after removing common 1st base) is:

B = "GGCTGGGCAGAGGCGCCCCTCACCTCCCG"
A  = "GACGGGGCGGCTGGCTGGGCGGGGAGCTGACCCCACCTCCCTCCCAGACGGGGCGGCTGGCCGGGCGGGGGGCTGACCCCCCCCCCACCTCCCTCCCAGACGGGGCGGCTGGCCGGGCGGGGGGCTGACCCCCCCCACCTCCCTCCCGGACGGGGCGTCTGGCCGGGTGGGGGGCTAACTCCCCCACCTCCCTCCTGGACGGGGTGGCTGGCGAGGCAGAGGGGCTCCTCACTTCCCAGTAGGGGA

[pd3]

I believe this can be closed? Please shout if not