Overlap number discrepancy between VCFTools and BEDTools

[zxyeo]

--------------------------------VCFTools-----------------------------------------------------
$./vcf-compare in.vcf.gz dbsnp_132.hg18.vcf.gz > in.snp-dbsnp.summary
$less in.snp-dbsnp.summary

Number of sites found only in
115 in.vcf.gz (0.8%)
15101 dbsnp_132.hg18.vcf.gz (0.1%) in.vcf.gz (99.2%)
26694223 dbsnp_132.hg18.vcf.gz (99.9%)

Number of REF matches: 15052
Number of ALT matches: 14449
Number of REF mismatches: 49
Number of ALT mismatches: 603
Number of sites lost due to grouping (e.g. duplicate sites)
2024327 (7.0%) .. read 28733651, reported 26709324 dbsnp_132.hg18.vcf.gz

--------------------------------BEDTools-----------------------------------------------------
$intersectBed -u -f 1 -a in.vcf -b dbsnp_132.hg18.vcf > in.snp-dbsnp.u.bed
$wc -l in.snp-dbsnp.u.bed
15092 in.snp-dbsnp.u.bed

***********************************************************************************************
As you can see, the number of overlap is different. It is 15101 from VCFTools but 15092 by BEDTools.

I also used the vcf-isec to get the VCFTools version of overlap vcf. Then, I used the 'intersectBed' to overlap the this vcf (VCFTools version) with the in.snp-dbsnp.u.bed (BEDTools version) and get all 15092 overlaps.

AFAIK, BEDTools did the overlap by only using the position information (i.e. consider an overlap even with different base(s)). But I am not sure what VCFTools does to come out with the additional 9 overlaps (or why BEDTools has 9 overlaps missing).

It would be great if someone could explain what VCFTools:vcf-isec is doing and give me some advice on how to interpret the above mentioned discrepancy. Many thanks!

[zxyeo]

Update: is BEDTools doing more than simply comparing the genomic position?

I just looked at the additional 9 entries and I would like to update my interpretation:

Let's start with 1 of the 9 entries (generally, other entries give the same observation):
chr2 79990299 . GAGC G 408.26 PASS AC=2;AF=1.00;AN=2;DP=12;FS=0.000;HRun=0;HaplotypeScore=63.0941;MQ=41.32;MQ0=0;QD=34.02;SB=-158.86;SF=0,1,1

In 'dbsnp_132.hg18.vcf', a similar entry was found:
chr2 79990299 rs10578220 G GAGC . PASS G5;G5A;GNO;NSF;REF;RSPOS=80136791;SAO=0;SCS=0;SLO;SSR=0;VC=INDEL;VP=050100001201030100000200;WGT=1;dbSNPBuildID=119

i.e. somehow my REF and ALT base(s) were switched.

Interestingly, BEDTools managed to distinguish them (which means BEDTools might be using the base identity on top of position information). I guess VCFTools might be only considering the position information (that's why this is detected as overlap). Correct me if this is not true.

[wanguan2000]

vcf-compare

Compares positions in two or more VCF files and outputs the numbers of positions contained in one but not the other files; two but not the other files, etc, which comes handy when generating Venn diagrams. The script also computes numbers such as nonreference discordance rates (including multiallelic sites), compares actual sequence (useful when comparing indels), etc.

vcf just considering the position information.