vibe-vep is a single Go binary with no Perl dependencies that predicts variant consequences using GENCODE annotations (~95MB download vs 17GB VEP cache).
It incorporates transcript prioritization similar to Genome Nexus — selecting a single gene and protein change by prioritizing coding transcripts and highest-impact consequences — while also providing effect predictions for all overlapping transcripts.
Achieves 99.8% concordance with GDC/VEP annotations across 1M+ TCGA variants.
Fast
~14,000 variants/sec parallel (4 workers), ~5,000 single-threaded. Cache loading in ~25 seconds for 254k GENCODE v46 transcripts.
Validated
Tested against 7 TCGA GDC studies (1,052,366 variants). 99.8% consequence match rate, 0 cancer gene mismatches.
Multiple Data Sources
AlphaMissense, ClinVar, Cancer Hotspots, SIGNAL, and OncoKB annotation sources. Extensible architecture for adding new sources.
Key Features
- VCF/MAF Parsing: Annotate variants from VCF or MAF files
- GENCODE Annotations: Uses GENCODE GTF/FASTA (~95MB download vs 17GB VEP cache)
- Consequence Prediction: Classifies variants using Sequence Ontology terms
- Validation Mode: Compare existing MAF annotations against predictions
- VCF to MAF Conversion: Convert VCF files to MAF format with annotations
- Annotation Sources: AlphaMissense, ClinVar, Cancer Hotspots, SIGNAL, OncoKB