ClawBio WorkshopVariant Interpretation
From a raw genome file to clinical findings
Dr Manuel Corpas · University of Westminster · 2026
Where we are
You have heard about ClawBio, how skills work, and the databases behind variant annotation. Now you will run the full pipeline yourself on a real genome.
The genome belongs to Dr Manuel Corpas, published under CC0 public domain in 2013. Everything you find is real.
This session: ~10 min slides, then ~20 min hands-on in Google Colab. By the end you will have annotated 21 clinical variants, identified five actionable findings, and generated a pharmacogenomics report.
Learning objectives
- Load and explore a consumer genotyping file (~600,000 SNPs)
- Select clinically relevant variants from a genome-wide dataset
- Annotate variants using Ensembl VEP, ClinVar, and gnomAD in one command
- Generate a pharmacogenomics report using CPIC guidelines
- Classify variants into priority tiers and explain clinical significance
- Run all analyses in Google Colab with zero infrastructure
Part 1
The data
What a genotyping file looks like
Consumer tests (23andMe, AncestryDNA) produce a text file with four columns:
| rsID | Chromosome | Position | Genotype |
|---|---|---|---|
| rs6025 | 1 | 169519049 | CT |
| rs1800562 | 6 | 26093141 | GA |
| rs113993960 | 7 | 117559590 | DI |
| rs9923231 | 16 | 31107689 | TT |
The 21 variants we will annotate
From 600,000 positions, ClawBio selects variants in genes with established clinical evidence:
Drug metabolism
CYP2C19, CYP2D6, CYP2C9, VKORC1, TPMT, DPYD, MTHFR
Disease risk
Factor V (clotting), CFTR (cystic fibrosis), HFE (iron), APOE (Alzheimer's)
Cancer predisposition
BRCA1, TP53
Each variant is converted to VCF format and sent to the Ensembl VEP REST API (free, public, no key required).
Part 2
What the pipeline returns
Priority tiers
ClawBio assigns every annotated variant to a tier based on clinical evidence and frequency:
| Tier | Definition | Example |
|---|---|---|
| Tier 1 | Pathogenic/likely pathogenic, rare (<0.1%), highest clinical relevance | CFTR deltaF508 (CF carrier) |
| Tier 2 | Drug response or established risk factor, actionable under CPIC | VKORC1 warfarin sensitivity |
| Tier 3 | VUS: insufficient evidence for classification | Rare missense without ClinVar entry |
| Tier 4 | Benign/likely benign, common (>1%) | MTHFR A1298C |
Reading the output table
The annotation report includes one row per variant with these columns:
| Column | What it tells you |
|---|---|
gene | Gene harbouring the variant |
consequence | Protein impact: missense, synonymous, frameshift |
impact | HIGH (protein disruption), MODERATE, LOW, MODIFIER |
clinvar_significance | Pathogenic, Likely pathogenic, VUS, Benign, Drug response |
gnomad_af | Global allele frequency. Below 0.001 (0.1%) = rare |
priority_tier | ClawBio tier 1-4 (see previous slide) |
Part 3
Six key findings
Factor V Leiden Tier 1
rs6025 in gene F5. Heterozygous carrier.
What it does
Makes blood clotting factor V resistant to inactivation. Blood clots more easily than it should.
Risk
3-8x increased risk of deep vein thrombosis (DVT). ~5% of Europeans carry this variant.
Clinical actions: Contraceptive counselling (oestrogen increases clot risk further). Perioperative planning. Long-haul flight precautions. Family testing.
Warfarin sensitivity Tier 2
Two genes, one dangerous interaction:
VKORC1 (rs9923231)
TT homozygous. Warfarin's target. TT genotype = high sensitivity. The drug works too well.
CYP2C9*2 (rs1799853)
Heterozygous carrier. Slows warfarin metabolism. The drug stays in the body longer.
Combined effect: CPIC recommends starting at 50-75% lower dose. Without genetic testing, a standard dose could cause a haemorrhage. This is the textbook example of pharmacogenomics saving lives.
Carrier findings Tier 1
CFTR deltaF508 (rs113993960)
Heterozygous carrier. Cystic fibrosis requires two copies. ~1 in 25 Europeans carry one mutation.
Action: Partner testing recommended. If both parents carry a mutation, 25% chance of affected child.
HFE C282Y (rs1800562)
Heterozygous carrier. Elevated iron absorption. Homozygotes develop hereditary haemochromatosis.
Action: Serum ferritin monitoring. Carriers virtually never develop disease. Homozygotes: blood donation regimen.
Risk factors and common variants
APOE e3/e4 (rs429358)
~3-fold increased late-onset Alzheimer's risk vs e3/e3. Two copies (e4/e4) = ~12-fold.
Probabilistic risk factor, not a diagnosis. Many carriers remain unaffected. Requires genetic counselling due to psychological impact.
MTHFR C677T (rs1801133) Tier 2
Heterozygous. ~65% enzyme activity. 30-40% of Europeans carry this.
One of the most over-interpreted variants in consumer genomics. Heterozygotes typically need no intervention. Homozygotes may benefit from methylfolate in pregnancy.
Part 4
Hands-on Practical
Open Google Colab now
What you will do
| Step | Task | Time |
|---|---|---|
| 0 | Setup: install ClawBio in Colab (two cells, ~15 seconds) | 2 min |
| 1 | Explore the Corpasome: load 600K SNPs, inspect chromosome distribution | 5 min |
| 2 | Select 21 clinically relevant variants and convert to VCF | 3 min |
| 3 | Annotate with VEP, ClinVar, gnomAD: one command, full results table | 5 min |
| 4 | Pharmacogenomics: CPIC drug-gene mapping and dosing recommendations | 5 min |
| 5 | Exercises: demo data, personal upload (optional), variant research | 10 min |
Requirements: A Google account and a web browser. Nothing to install. No API keys. No payment.
Exercises
| Exercise | Task | |
|---|---|---|
| 5a | Rerun the pipeline on 20 synthetic variants using the --demo flag. Compare your output to the expected results. | Required |
| 5b | Upload your own 23andMe or AncestryDNA file (if you have one) and run the pipeline on your data. | Optional |
| 5c | Pick one gene from the output. Research: what is its ACMG classification? How common is it? Would you report it? | Required |
What SNP arrays miss
Consumer genotyping arrays test ~600,000 positions out of 3 billion in the genome.
Not detected
- Structural variants (deletions, duplications, inversions)
- Most rare variants (private to individual or family)
- Copy number changes
- Short insertions/deletions in most regions
Detected
- Common SNPs at pre-selected positions
- Well-studied pharmacogenomic variants
- Established disease-risk variants
- Ancestry-informative markers
A "clear" result does not mean the genome is free of pathogenic variants. It means the array did not test the position where a pathogenic variant might exist.
Take-home messages
- 21 variants, five actionable findings. A real genome with real clinical implications, analysed in minutes.
- Pharmacogenomics prevents harm today. The warfarin double-hit in this genome would change a prescription right now.
- Carrier status matters for family planning. CFTR and HFE findings are not about the individual alone.
- "We don't know yet" is honest. Over half of all variants are VUS. Resist the pressure to over-interpret.
- SNP arrays have blind spots. A negative result is not a clean bill of health. 30x WGS sees far more.
Resources
| Resource | Link |
|---|---|
| ClawBio GitHub | github.com/ClawBio/ClawBio |
| Tutorial page | docs.clawbio.ai/tutorials/variant-interpretation-workshop |
| Ensembl VEP | ensembl.org/vep |
| ClinVar | ncbi.nlm.nih.gov/clinvar |
| gnomAD | gnomad.broadinstitute.org |
| CPIC guidelines | cpicpgx.org |
| Corpasome (Zenodo) | doi:10.5281/zenodo.19297389 |
| GWAS session (next) | docs.clawbio.ai/tutorials/gwas-workshop |