Personal-Genome-Pipeline

Step 21: CYP2D6 Star Allele Calling with Cyrius

EXPERIMENTAL: Cyrius is installed via pip at runtime inside a generic Python container, which is fragile (network dependency, version drift). Results should be cross-referenced with PharmCAT’s CYP2D6 call.

What This Does

Calls CYP2D6 star alleles (diplotypes) from your WGS BAM using Illumina’s Cyrius tool. CYP2D6 is the single most important pharmacogene — it metabolizes roughly 25% of clinically used drugs — but its highly homologous pseudogene (CYP2D7) and frequent structural rearrangements (deletions, duplications, gene-pseudogene hybrids) make it extremely difficult to genotype from short reads.

Why

PharmCAT (step 7) handles most pharmacogenes well, but its internal CYP2D6 calling is limited for WGS data. Cyrius was purpose-built by Illumina to resolve CYP2D6 using read-depth patterns across the CYP2D6/CYP2D7 region. Running Cyrius separately gives you a CYP2D6 diplotype that you can cross-reference with PharmCAT’s results.

Tool

Cyrius (Chen et al., Pharmacogenomics J 2021) – Illumina’s depth-based CYP2D6 caller

Docker Image

python:3.11-slim

Cyrius is installed via pip install cyrius inside the container at runtime. No dedicated Cyrius Docker image is required.

Input

Sorted BAM with index from alignment (step 2):
- ${GENOME_DIR}/${SAMPLE}/aligned/${SAMPLE}_sorted.bam
- ${GENOME_DIR}/${SAMPLE}/aligned/${SAMPLE}_sorted.bam.bai

Command

./scripts/21-cyrius.sh your_name

What the Script Does Internally

Validates that the sorted BAM and its index exist
Creates a manifest file listing the BAM path (Cyrius requires this)
Installs Cyrius in a Python 3.11 container and runs star_caller with --genome 38 (GRCh38)
Parses the output TSV to display the called diplotype

Output

File	Contents
`${SAMPLE}_cyp2d6.tsv`	Tab-delimited results with sample name, diplotype, and supporting evidence

All output is written to ${GENOME_DIR}/${SAMPLE}/cyrius/.

Runtime

~5-15 minutes (includes pip install overhead on first run).

Interpreting Results

The output TSV contains the CYP2D6 diplotype in star-allele notation, for example:

Sample	Genotype
sergio	1/2

Common results and what they mean:

*1/*1 – Normal metabolizer (two fully functional copies)
*1/*2 – Normal metabolizer (*2 is also functional)
*1/*4 – Intermediate metabolizer (*4 is non-functional)
*4/*4 – Poor metabolizer (no functional copies)
*1/*1xN – Ultrarapid metabolizer (gene duplication, N extra copies)
*5/*5 – Poor metabolizer (whole gene deletion)

Look up your specific diplotype at PharmGKB CYP2D6 for the corresponding metabolizer phenotype and drug implications.

Drugs affected by CYP2D6 status

Codeine, tramadol, oxycodone, tamoxifen, ondansetron, atomoxetine, most tricyclic antidepressants (amitriptyline, nortriptyline), and paroxetine – among many others.

Limitations

Cyrius works best with 30X+ WGS data. Lower coverage may produce uncertain calls.
Rare hybrid alleles (e.g., *36, *68) may not be resolved.
The pip-install-at-runtime approach adds startup time and requires internet. If you run this frequently, consider building a custom Docker image with Cyrius pre-installed.
Cyrius only calls CYP2D6. For other pharmacogenes, rely on PharmCAT (step 7).
Cyrius has not been updated since May 2021 (v1.1.1). A 2025 study (BCyrius, PMID 39901590) found Cyrius fails to call or miscalls 50/360 simulated samples (13.9%) due to its outdated star allele database. Consider Aldy as an alternative (see below).

Recommended Alternative: Aldy

Aldy v4.8.3 is a leading CYP2D6 caller for short-read WGS data. A systematic comparison (Twesigomwe et al. 2020, PMID 32789024) found Aldy was “the best performing algorithm in calling CYP2D6 structural variants.” It identifies 92.2% of currently defined minor star alleles (vs 85.6% for Cyrius) and is actively maintained with the current PharmVar database.

Aldy also calls 37 additional pharmacogenes (CYP2C19, CYP2B6, UGT1A1, NAT2, DPYD, SLCO1B1, etc.), which can supplement PharmCAT results.

To use Aldy instead of Cyrius:

# Install in a Python container (one-time, or build a custom image)
docker run --rm --user root \
  -v ${GENOME_DIR}:/genome \
  python:3.11-slim \
  bash -c "
    pip install -q aldy==4.8.3 &&
    aldy genotype \
      -p illumina \
      -g CYP2D6 \
      -o /genome/${SAMPLE}/cyrius/${SAMPLE}_aldy_cyp2d6.aldy \
      /genome/${SAMPLE}/aligned/${SAMPLE}_sorted.bam
  "

License note: Aldy uses an academic/non-commercial license (IURTC, Indiana University). It is free for personal and research use but is NOT compatible with GPL-3.0 redistribution. This is why it is documented here as an optional recommendation rather than replacing Cyrius in the pipeline script. A GPL-compatible alternative (pypgx) is available in step 32.

Notes

The script creates a manifest file listing the BAM path, then runs Cyrius in a single container invocation.
Cross-reference the Cyrius diplotype with PharmCAT’s CYP2D6 call and pypgx (step 32). Cyrius can fail on some WGS samples due to CYP2D7 pseudogene homology; pypgx handles this more robustly. If all three disagree, Aldy (see above) has the broadest star allele coverage among available callers.
Step 27 (CPIC lookup) currently reads PharmCAT JSON only. To translate a Cyrius diplotype into drug recommendations, consult CPIC guidelines manually.