bcftools icon indicating copy to clipboard operation
bcftools copied to clipboard
verified publisher icon samtools

Is this expected bcftools norm behaviour for joining biallelics into a multiallelic?

Open sajeevbatra opened this issue 2 years ago • 1 comments

Example with multiallelic: cat test20.vcf ##fileformat=VCFv4.2 ##FORMAT=<ID=GT,Number=1,Type=String,Description="Genotype"> ##contig=<ID=chr7> #CHROM POS ID REF ALT QUAL FILTER INFO FORMAT SAMPLE1 chr7 117559591 Indel TCTT T,TTCT . . . GT 0/0

Step1: bcftools norm -m-any -f ../hg38/hg38.fa test20.vcf Splitting works OK.

###fileformat=VCFv4.2 ##FILTER=<ID=PASS,Description="All filters passed"> ##FORMAT=<ID=GT,Number=1,Type=String,Description="Genotype"> ##contig=<ID=chr7> ##bcftools_normVersion=1.17+htslib-1.17 ##bcftools_normCommand=norm -m-any -f ../hg38/hg38.fa test20.vcf; Date=Sat Mar 11 15:36:30 2023 #CHROM POS ID REF ALT QUAL FILTER INFO FORMAT SAMPLE1 chr7 117559590 Indel ATCT A . . . GT 0/0 chr7 117559592 Indel CT TC . . . GT 0/0 Lines total/split/realigned/skipped: 1/1/2/0

Step2: Splitting and joining but joining does not work. Why?

bcftools norm -m-any -f ../hg38/hg38.fa test20.vcf | bcftools norm -m+any -f ../hg38/hg38.fa Lines total/split/realigned/skipped: 1/1/2/0 ##fileformat=VCFv4.2 ##FILTER=<ID=PASS,Description="All filters passed"> ##FORMAT=<ID=GT,Number=1,Type=String,Description="Genotype"> ##contig=<ID=chr7> ##bcftools_normVersion=1.17+htslib-1.17 ##bcftools_normCommand=norm -m-any -f ../hg38/hg38.fa test20.vcf; Date=Sat Mar 11 15:37:29 2023 ##bcftools_normCommand=norm -m+any -f ../hg38/hg38.fa; Date=Sat Mar 11 15:37:29 2023 #CHROM POS ID REF ALT QUAL FILTER INFO FORMAT SAMPLE1 chr7 117559590 Indel ATCT A . . . GT 0/0 chr7 117559592 Indel CT TC . . . GT 0/0 Lines total/split/realigned/skipped: 2/0/0/0

The above command correctly splits my multiallelic record into 2 biallelic records and then it should join them into a single multiallelic record but it does not. Am I doing something wrong or is this an issue with bcftools norm? Thanks.

sajeevbatra avatar Mar 13 '23 05:03 sajeevbatra

This is the expected behavior for now, the program is not smart enough to join the two records split across different positions. The inverse of bcftools norm --atomize is needed for this.

Just to cross-reference related issues, see also https://github.com/samtools/bcftools/issues/327, https://github.com/samtools/bcftools/issues/1724, https://github.com/samtools/bcftools/issues/1843

pd3 avatar Mar 20 '23 11:03 pd3