Manual variant annotation is a multi-step process that most genomics labs still do by hand: ClinVar lookups, gnomAD frequency checks, HGVS notation, gene panel filtering. For a 200-variant exome, that's 3–6 hours of work. Baseshift automates the entire pipeline. This post explains exactly how.
The problem with manual variant annotation
A typical clinical genomics lab receives an exome VCF with 50–300 variants that need review before a report goes to a clinician. The annotation workflow looks something like this:
- Export variants to a spreadsheet or local tool
- Query ClinVar by rsID for pathogenicity classification — variant by variant
- Check gnomAD allele frequencies via the web browser or API
- Convert genomic coordinates to HGVS notation by hand
- Cross-reference gene names against gene panels (OMIM, PanelApp)
- Flag variants of uncertain significance, escalate to senior staff
- Export findings into a clinical report document
For a single exome, this is 3–6 analyst hours. At a lab running 30–50 exomes per week, that's 90–300 analyst-hours per week consumed by annotation work — time that could go toward interpretation, case review, and higher-value clinical decisions.
The tools exist to automate this. The bottleneck is that most labs stitching together 4–5 disconnected tools (ClinVar web, gnomAD browser, an annotation script, a reporting template) rather than using an integrated pipeline.
Baseshift's approach: Upload a VCF once. Get a complete annotated variant table and a clinical PDF report in under 60 minutes. No spreadsheet juggling, no manual ClinVar lookups, no copy-paste into a report template.
The Baseshift annotation pipeline
Here's what happens from the moment you upload a VCF to when your report is ready:
VCF Parsing
The uploaded VCF is parsed for chromosome, position, reference allele, alternate allele, quality scores, and FILTER flags. Each variant is normalized (left-aligned, trimmed) to ensure consistent coordinates before annotation.
ClinVar Cross-Reference
Each variant's genomic coordinates (chr:pos:ref:alt) are matched against ClinVar's latest release. The response includes VCV accession, clinical significance (benign/pathogenic/VUS), conditions, and review status (star rating). Results are cached to avoid repeated lookups for the same variant.
gnomAD Frequency Lookup
gnomAD v4 allele frequencies are fetched for each variant using the gnomAD GraphQL API. Population-level AF values are displayed alongside the ClinVar classification to flag variants that are common in population databases — a strong signal against clinical significance.
HGVS Notation + Gene Annotation
HGVS genomic notation (c. and p. syntax) is generated from the VCF coordinates. Gene names are mapped to RefSeq and Ensembl identifiers. Gene panels from PanelApp are cross-referenced to flag variants in relevant clinical panels (developmental disorders, oncology, cardiomyopathy).
Clinical Report PDF Generation
A structured PDF report is generated containing the full annotated variant table, ClinVar significance badges, gnomAD frequency tiers, gene panel matches, and a clinical summary section. The report is ready for clinician review and sign-off.
What's in a Baseshift report
The variant table is the core of the report. Every row shows:
| Variant | Gene | ClinVar Significance | gnomAD Freq | ZYG | HGVS |
|---|---|---|---|---|---|
| chr7:g.117171655C>T | CFTR | Pathogenic | 0.0002 | Hom | c.1657C>T (p.Arg553*) |
| chr12:g.25398280C>T | PKP2 | Likely Benign | 0.012 | Het | c.2185C>T (p.Arg729*) |
| chr3:g.36993341T>C | MLH1 | VUS | 0.0008 | Het | c.1958T>C (p.Met653Thr) |
The color-coded badges make it immediately obvious which variants need attention. Pathogenic variants with low gnomAD frequency and a strong ClinVar classification bubble to the top. Common variants with high population frequency are deprioritized. VUS variants are flagged for interpretation by a senior analyst.
The report also supports gene panel filtering — you can filter the variant table to show only variants matching a specific clinical panel (e.g., "Developmental Disorders GMS" or "Hereditary Cancer"). This is how labs reduce a 200-variant list to the 8–12 variants that are actually relevant to the referral question.
Try it with your own VCF
The interactive demo at /demo walks through the full Baseshift pipeline without requiring an account — upload a VCF, watch the analysis run, inspect the variant table, and preview the PDF export.
For the full experience with your own data, sign up at /signup. First analysis is free, no credit card required.
Run your first variant analysis free
Upload a VCF, get an annotated clinical report in under 60 minutes. No account required for the demo — or sign up to process your own data.
See the interactive demo → Create free account $10 per exome after free trial · 50% off for first 100 samples