GRCh37:17:41197726:AG:A is predicted as stop_gained for NM_007294.4 but reported as frameshift

[holtgrewe]

Describe the bug Put the variant GRCh37:17:41197726:AG:A as VCF. Annotate with SnpEff and GRCh37.p13. See that it is predicted to be stop_gained for NM_007294.4. Compared with VariantValidator, it should be NP_009225.1:p.(Leu1854Ter).

To Reproduce

1. SnpEff version: 5.1d
2. Genome version: GRCh37.p13
3. SnpEff full command line: java -jar snpEff.jar -v GRCh37.p13 bug.in.vcf >bug.out.vcf
4. Output / Error message: SnpEff succeeds, no error message.

Expected behavior Prediction should be stop gained, as in VariantValidator output:

Data

• bug.in.vcf.txt
• bug.out.vcf.txt

Additional context N/A

[holtgrewe]

The same is true for

• GRCh37:17:41197789:AC:A
• GRCh37:17:41197801:TCA:T

[holtgrewe]

Something weird is going on for

• GRCh37:17:41201181:CCACA:TCACT

as well. SnpEff predicts as missense_variant but should be disruptive_inframe_deletion, I believe:

[holtgrewe]

The variant

• GRCh37:17:41203121:AG:A

is predicted to be frameshift_variant but should probably be predicted as stop_gained

(Note that all the variants from this issue are from ClinVar for BRCA1, so they are clinically relevant; although frameshift and stop_gained are both nonsense).

[holtgrewe]

Same for GCh37:17:41203122:GC:G

[holtgrewe]

Same for GRCh37:17:41209143:C:CA