END less than start in gnomad SV VCF?

[cmdcolin]

Hi there
I work on a genome browser where we'd like to helpfully display SVs but the code doesn't like when it sees features where END<START

I saw in the SV VCF that there is a line where the "END" field is less than the start

https://storage.googleapis.com/gnomad-public/papers/2019-sv/gnomad_v2_sv.sites.vcf.gz

1 1386372 gnomAD_v2_INS_1_51      N       <INS>   372     PASS    END=1386324;SVTYPE=INS;CHR2=1;SVLEN=78;ALGORITHMS=delly,mant
a;EVIDENCE=SR;SOURCE=INS_1:1451663-1451741;PROTEIN_CODING__INTRONIC=ATAD3C;AN=20020;AC=143;AF=0.007143;N_BI_GENOS=10010;N_HOMREF=9867;N_HET=143;N_HOMALT=0;FREQ_HOMREF=0.985714;FREQ_HET=0.0142857;FREQ_HOMALT=0;AFR_AN=8062;AFR_AC=138;AFR_AF=0.017117;AFR_N_BI_GENOS=4031;AFR_N_HOMREF=3893;AFR_N_HET=138;AFR_N_HOMALT=0;AFR_FREQ_HOMREF=0.965765;AFR_FREQ_HET=0.0342347;AFR_FREQ_HOMALT=0;AMR_AN=1766;AMR_AC=3;AMR_AF
=0.001699;AMR_N_BI_GENOS=883;AMR_N_HOMREF=880;AMR_N_HET=3;AMR_N_HOMALT=0;AMR_FREQ_HOMREF=0.996602;AMR_FREQ_HET=0.00339751;AMR_FREQ_HOMALT=0;EAS_AN=2222;EAS_AC=0;EAS_AF=0;EAS_N_BI_GENOS=1111;EAS_N_HOMREF=1
111;EAS_N_HET=0;EAS_N_HOMALT=0;EAS_FREQ_HOMREF=1;EAS_FREQ_HET=0;EAS_FREQ_HOMALT=0;EUR_AN=7586;EUR_AC=1;EUR_AF=0.000132;EUR_N_BI_GENOS=3793;EUR_N_HOMREF=3792;EUR_N_HET=1;EUR_N_HOMALT=0;EUR_FREQ_HOMREF=0.99
9736;EUR_FREQ_HET=0.000263644;EUR_FREQ_HOMALT=0;OTH_AN=384;OTH_AC=1;OTH_AF=0.002604;OTH_N_BI_GENOS=192;OTH_N_HOMREF=191;OTH_N_HET=1;OTH_N_HOMALT=0;OTH_FREQ_HOMREF=0.994792;OTH_FREQ_HET=0.00520833;OTH_FREQ
_HOMALT=0;POPMAX_AF=0.017117

I was wondering if this has any meaning or maybe it is a liftover or data bug? Here is the link in the browser https://gnomad.broadinstitute.org/variant/INS_1_35?dataset=gnomad_sv_r2_1

[nawatts]

Hi @cmdcolin,

I believe this is related to talkowski-lab/gnomad-sv-pipeline#3 (@RCollins13 would have to confirm that).

The SV dataset shown in the browser has been updated from the VCF linked in the issue description. The SV dataset shown in the browser comes from https://storage.googleapis.com/gnomad-public/papers/2019-sv/gnomad_v2.1_sv.sites.vcf.gz. In the 2.1 VCF, end is greater than start for the variant "gnomAD-SV_v2.1_INS_1_35".

1	1386372	gnomAD-SV_v2.1_INS_1_35	N	<INS>	692	UNSTABLE_AF_PCRMINUS	END=1386373;SVTYPE=INS;SVLEN=78;CHR2=1;POS2=1386324;END2=1386325;ALGORITHMS=delly,manta;EVIDENCE=SR;HIGH_SR_BACKGROUND;PROTEIN_CODING__INTRONIC=ATAD3C;SOURCE=INS_1:1451663-1451741;AN=18980;AC=29;AF=0.001528;N_BI_GENOS=9490;N_HOMREF=9461;N_HET=29;N_HOMALT=0;FREQ_HOMREF=0.996944;FREQ_HET=0.00305585;FREQ_HOMALT=0;AFR_AN=7500;AFR_AC=29;AFR_AF=0.003867;AFR_N_BI_GENOS=3750;AFR_N_HOMREF=3721;AFR_N_HET=29;AFR_N_HOMALT=0;AFR_FREQ_HOMREF=0.992267;AFR_FREQ_HET=0.00773333;AFR_FREQ_HOMALT=0;AMR_AN=1444;AMR_AC=0;AMR_AF=0;AMR_N_BI_GENOS=722;AMR_N_HOMREF=722;AMR_N_HET=0;AMR_N_HOMALT=0;AMR_FREQ_HOMREF=1;AMR_FREQ_HET=0;AMR_FREQ_HOMALT=0;EAS_AN=2410;EAS_AC=0;EAS_AF=0;EAS_N_BI_GENOS=1205;EAS_N_HOMREF=1205;EAS_N_HET=0;EAS_N_HOMALT=0;EAS_FREQ_HOMREF=1;EAS_FREQ_HET=0;EAS_FREQ_HOMALT=0;EUR_AN=7440;EUR_AC=0;EUR_AF=0;EUR_N_BI_GENOS=3720;EUR_N_HOMREF=3720;EUR_N_HET=0;EUR_N_HOMALT=0;EUR_FREQ_HOMREF=1;EUR_FREQ_HET=0;EUR_FREQ_HOMALT=0;OTH_AN=186;OTH_AC=0;OTH_AF=0;OTH_N_BI_GENOS=93;OTH_N_HOMREF=93;OTH_N_HET=0;OTH_N_HOMALT=0;OTH_FREQ_HOMREF=1;OTH_FREQ_HET=0;OTH_FREQ_HOMALT=0;POPMAX_AF=0.003867

[cmdcolin]

Ah didn't realize this is a subset of that problem, I saw this in similar context in a Sniffles VCF too. Thanks for clarifying

[RCollins13]

Hi @cmdcolin,

This is an issue with several SV algorithms, and one we're actively working to resolve for gnomAD.

The v2.1 SV release (available here: https://gnomad.broadinstitute.org/downloads) should have these issues corrected, and future gnomAD releases will also not include records where END<POS.

Thanks,
Ryan