bcftools view -r seems to be getting sites outside of the region I designate?

Question

[curious]

my_vcf.vcf:

##fileformat=VCFv4.1
#CHROM  POS     ID      REF     ALT     QUAL    FILTER  INFO    FORMAT  SMAPLE
chr22   50482719        chr22:50482719:AACACTCGGGCCCCCGAAG:A    AACACTCGGGCCCCCGAAG     A       .       PASS    AF=7e-05     GT    0|0
chr22   50482725        chr22:50482725:CGGGCCCCCGAAGACA:C       CGGGCCCCCGAAGACA        C       .       PASS    AF=7e-05     GT    0|0
chr22   50482740        chr22:50482740:A:C      A       C       .       PASS    AF=7e-05     GT    0|0

bcftools view -r chr22:50482735-50487454 my_vcf.vcf returns all these rows, not just the last one. Is this a bug or am I missing something obvious. I seem to be using the -r switch correctly: -r, --regions chr|chr:pos|chr:beg-end|chr:beg-[,…]

The only other thing that I can think of is maybe bcftools does something internally to adjust the position of indels, if that is the case it really hurts my workflow :(

Answer 1

[Ram]

You are correct - bcftools returns all positions overlapping the region specified in -r. The first two entries' REF allele overlaps your region by at least one base, so they're returned. What happens when you use -t instead of -r?

[curious]

bcftools view -r chr22:50482735-50487454 -t chr22:50482735-50487454 my_vcf.vcf

Looks like I can use them both to avoid this issue and still benefit from tab indexing.

just noticed in docs:

Yet another difference between the two is that -r checks both start and end positions of indels, whereas -t checks start positions only.