norm doesn't normalize symbolic variants

[davmlaw]

Hi, thanks for bcftools!

A variant NC_000003.11:g.128204049_128206714del can be written in a VCF as having an explicit REF or via a symbolic alt

bcftools norm will left-align the former (to position 128204042), but not the symbolic representation - I think it should normalize both

#CHROM	POS	ID	REF	ALT	QUAL	FILTER	INFO
NC_000003.11	128204048	.	GAGGAGATCAGGGAGCCATCGAAATCCCAAGATCAGACTGATTGAGTTAGAGACCCAGATTCCTTAATGGTCTGGAACCTCCGGAGTGCCTGAAACATGCACACACAACATGCACACAGACACACGTATACACATGCACACACGCTCCCAAACACATGCACATACAGAGTCACTTCCCTCGCTTCACATACTAAGTCCTGAGGTGGCTTGAATTTTCACTTAAGATTCAGGGAGGGAAAGGGGGAATTCCTGTTCTATGTTCTGGTCAGGCAGTGACCAACCCTGGGGCAAGGAACTGAACTTTGGGGGTACACTGGAAGCACTTAAGAAAATGGCAAAAGTTTTAGAGTCTCCTCTCCCTGACCCGGGGGTCTCAAACATCTGCTGGGGGCTATTAGAGCGAGACATCACCCATCCCCAGATCTGGGAAACCAACACTGCCACCTCTCCCAAGTCACAGCTCCCCACCACAAAAACGCAAATGCTCCCCTCTTCCACGAAGTCCCCAGCACCTGCCTTTACCTGAACAGGAACGAGCCTTGCTGCGCTGCTTAGGGGTGAAGCTGGAGGCCGGTCCCCCCAGGAAGCCTCCGGGGTGGAAGAGTCCGCTGCTGTAGTCGTGGGCAGCCGCCGGCACATAGGAGGGGTAGGTGGGGATGGGGTGGTGTGTAGCAGGCTGGGTGCCCATAGTAGCTAGGCCTGGGCGCAGGGGACTGCCACTTTCCATCTTCATGCTCTCCGTCAGTGACACCTGGTACTTGACGCCGTCCTTGTCCTCTCCTCGGGCTGCACTACCCCCCGCGGAAGATGAGGCTGGAGACGCAGCCCCCGTGGTGCTAGGGTCAGGAGACACTTCTTTGGGTGGCGTGGGTGGGAAGCCGAAAAGGTGGGAGCCAGAGTGGGCTGCTGTAGGGGTGAGGGAGGCCACTGAGCTCCCGCTGCCTCCCCCGCTCCCACCCCCAGCCCCTGGGTACACAGAGAGTGGGCCTCCAGGGCCTCCAGCAGCTGAGGGGTGCAGTGGCGTCTTGGAGAAGGGGCTCACGGTCCAGGGGTTGTGGTGGTGGGCCGCAGCGGCAGAGAGGGCTGCTTTGCCCCCGTCCAGCCAGGGCAAACCCGGGCTGTGCAACAAGTGTGGGCGGCACATCTGGCCTCCGGTCAGGCGGGCTGCGGGCAAAGAGAGAGAGGATCAGGGTGGGCAGAAAGATCAGGGTAGGCAGAGCTAGGGACGCCCCTGACAGACATTGAGATCACGACTCCCAGAACCAGCAGTCATCCCCTCCCCAAAGAAAGCCAGAAACATAATACCCCACCGGTAATAATCAGGAATGTCAGTCCAAGCTGAAGGACAAGTGGCATAGAAGGAACCCCACCGGACAGACCCTACAGGGAACCCTCACAGGCCAGCTGGAAGTGGGCAGAAACCCTGTGGGTCCCAGACCCTCCCCAATCGGCCGCTGCTCCCACCTCTCCCGCCCCAATTTTTCAGCAGCTCGATTCCTGCGGATCCTACATCCGGGAAGCAAGCAGACGGGCCCTCCTCCCCTCCCTCGCCTGGCGCGCGGCGCCTGGGTTCTCATCACCACGGGCCCAGTGCTCACCGTGCGCGGGGCTGTAGGAGACGCGCGCCCGCGCGTGAGCGGGGTTGGCATAGTAGGGGTTGCCCTGCGAGTCGAGGTGATTGAAGAAGACGTCCACCTCGTCTGGAGGCAGCAGCTGCGCGGGTTCCATGTAGTTGTGCGCCAGGCCCGGGTGGTGTGAGTCGGGGTGCTGCGCATTCAGCACGGCCGGGTGCGCCATCCAGCGCGGCTGCTCGGGCGCCACCTCCATGGCCGGCGGCGGCGGCTCAGGGTCTGGGTGCAGACGGCAACGGCCCTGCGCGAGGAAGGGGGAGTGAGGCGTGCCGCCAGCGCCTGACACCCCCCAAAGTCCCACCACGAGGTGTCCCGCACGCCACGGAGCCCCAGCCCAGATCCGGCGAGAAAGAGCACCAGTCCCGGGTGGGAGGAAAGCCCAAGGCTCAAAACGAAAGGAAGGCGGGGGAGGGGGTTCAGCCACGCACACTCACGTGGTGACCCGCGGCTCCAGAATCACACACCCGTGCACATGGGGTCACGCCCGGGGACGGGTCCCGACACCAGTGACCCCAACAAACGCACAGAGCAGCACTTCAGTCAGACACTCACACTGAGCCCCCCCGCCCGGTAGACAAACACATGAACACAGACTCAAAAGTTGGAGACAGGCGCCCGGGCACCCAGTGTGGCACTTGATCCCAGCGACACGCACACACCCACACTTGGCGCCAGATACACATACTGATCTCAACCCCGAAAACATGCACACGCAGCCCCCTGAGCGCAGTACTAAGCGGCACAATCAGGACCTCTCAACAAAGCACACCAAAGCAGTCGCCCGCAGCCTGGCCCCCCGCCCTAAGTCCCCCCAGAGTCCCCTCAAAGCTAGGAGCGCCCCAGGCCCCCAGCCGGCTCTCAAACCCCAAACTTACACACGCAGCCGTGGGGAGGGGAGGGACTCGGCCTCTGAGAGTGAAGGAGTTCCGGCGGGAGCCCCGAGGGCGACGGGCCCAGGGACAGCACGTCCGGAGGCTGGCGGGGCTTACAGGGTAGGAGCTGGGGGTAGAGTGCGCCTCGGCCTCGGGCCCTCCCG	G	.	PASS	.
NC_000003.11	128204048	.	G	<DEL>	.	PASS	END=128206714;SVTYPE=DEL;SVLEN=-2666

I have attached this as a vcf (with .txt extension)
indel_normalise_test.GRCh37.txt

[davmlaw]

In case I end up attempting to fix this (warning - I haven't written C in 15 years) I am interested in what people believe a good fix would be. Perhaps the way to solve this is to add an option to convert from symbolic variants to explicit ones, and back again. Ie you could go:

bcftools cmd --symbolic-to-explicit | bcftools norm | bcftools cmd --explicit-to-symbolic --max-allele-length=1000

I am not sure what command to put this under - am using "cmd", though perhaps --symbolic-to-explicit could be part of the norm command

[pd3]

An experimental support for normalizing symbolic alleles has been added. So far it only considers deletions (<DEL.*>) but nothing else