filtering expression - handling of missing values by != and !~

[aheinzel]

Dear all,

I have a question regarding the handling of missing values when using != and !~ in filtering expression in bcftools v1.19+htslib-1.19. To support my question I include a minimal vcf file with one info tag named TAG.

##fileformat=VCFv4.2
##FILTER=<ID=PASS,Description="All filters passed">
##contig=<ID=chr1,length=5>
##INFO=<ID=TAG,Number=.,Type=String,Description="Some tag">
##FORMAT=<ID=GT,Number=1,Type=String,Description="Genotype">
##bcftools_viewVersion=1.19+htslib-1.19
#CHROM	POS	ID	REF	ALT	QUAL	FILTER	INFO	FORMAT	S1
chr1	1	.	*	*	.	.	TAG=a,b,c	GT	0/0
chr1	2	.	*	*	.	.	TAG=a	GT	0/0
chr1	3	.	*	*	.	.	TAG=.	GT	0/.
chr1	4	.	*	*	.	.	TAG=.,.	GT	./0
chr1	5	.	*	*	.	.	TAG=a,.,c	GT	./.

I can use the standard string comparison operator in view -i TAG[*]!="." to include only sites with at least one non missing value for TAG:

##fileformat=VCFv4.2
##FILTER=<ID=PASS,Description="All filters passed">
##contig=<ID=chr1,length=5>
##INFO=<ID=TAG,Number=.,Type=String,Description="Some tag">
##FORMAT=<ID=GT,Number=1,Type=String,Description="Genotype">
##bcftools_viewVersion=1.19+htslib-1.19
##bcftools_viewCommand=view -i TAG[*]!="." input.vcf; Date=Sat Jan 18 17:33:45 2025
#CHROM	POS	ID	REF	ALT	QUAL	FILTER	INFO	FORMAT	S1
chr1	1	.	*	*	.	.	TAG=a,b,c	GT	0/0
chr1	2	.	*	*	.	.	TAG=a	GT	0/0
chr1	5	.	*	*	.	.	TAG=a,.,c	GT	./.

Doing the same with the regex operator view -i TAG[*]!~"\." does not filter out any variants:

##fileformat=VCFv4.2
##FILTER=<ID=PASS,Description="All filters passed">
##contig=<ID=chr1,length=5>
##INFO=<ID=TAG,Number=.,Type=String,Description="Some tag">
##FORMAT=<ID=GT,Number=1,Type=String,Description="Genotype">
##bcftools_viewVersion=1.19+htslib-1.19
##bcftools_viewCommand=view -i TAG[*]!~"\." input.vcf; Date=Sat Jan 18 17:37:27 2025
#CHROM	POS	ID	REF	ALT	QUAL	FILTER	INFO	FORMAT	S1
chr1	1	.	*	*	.	.	TAG=a,b,c	GT	0/0
chr1	2	.	*	*	.	.	TAG=a	GT	0/0
chr1	3	.	*	*	.	.	TAG=.	GT	0/.
chr1	4	.	*	*	.	.	TAG=.,.	GT	./0
chr1	5	.	*	*	.	.	TAG=a,.,c	GT	./.

Sorry just realized that this is maybe non conclusive, thus I add another example with view -i TAG[*]!~"[A-z]"

##fileformat=VCFv4.2
##FILTER=<ID=PASS,Description="All filters passed">
##contig=<ID=chr1,length=5>
##INFO=<ID=TAG,Number=.,Type=String,Description="Some tag">
##FORMAT=<ID=GT,Number=1,Type=String,Description="Genotype">
##bcftools_viewVersion=1.19+htslib-1.19
##bcftools_viewCommand=view -i TAG[*]!~"[A-z]" input.vcf; Date=Sun Jan 19 15:15:24 2025
#CHROM	POS	ID	REF	ALT	QUAL	FILTER	INFO	FORMAT	S1
chr1	3	.	*	*	.	.	TAG=.	GT	0/.
chr1	4	.	*	*	.	.	TAG=.,.	GT	./0
chr1	5	.	*	*	.	.	TAG=a,.,c	GT	./.

Does the negated regex operator automatically evaluate to true for missing values?

[pd3]

Thank you for the issue. This has been corrected so that one gets the expected results