Pharmacogenetics

Each star allele has a function assignment (no, decreased, normal or increased activity). The two inherited alleles yield an activity score and from it the metaboliser class. A poor metaboliser can barely activate a prodrug such as clopidogrel, an ultrarapid metaboliser converts codeine to morphine too quickly. What matters is not the single SNP but the haplotype: only the phase of the variants on the same chromosome defines the functional allele.

The pharmacogenetically most important genes sit in difficult genome regions. CYP2D6 borders the near-identical pseudogenes CYP2D7 and CYP2D8, so short reads map ambiguously and neighbouring sequence is wrongly taken over. Added to this are gene duplications, full deletions and hybrid alleles assembled from parts of gene and pseudogene. A plain variant list is not enough here: without phasing and copy number it stays unclear which variants sit on which allele and how many copies are active.

Genome combines Aldy 4 and PharmCAT. Aldy frames star-allele calling as an optimisation problem and solves it with integer linear programming: from the observed read depth and SNV evidence it first infers copy number and structural configuration, then decomposes this into the most likely pair of known alleles (allelic decomposition). This lets it genotype even structurally variable genes such as CYP2D6 with duplications and hybrids directly from the BAM. PharmCAT takes the alleles and diplotypes, fills in the remaining PGx genes from a standard VCF and translates the result into CPIC and DPWG guideline recommendations. The pharmacogenetics report is strictly separated from medical genomics and carries evidence, tool versions and provenance.

The technical typing ends at the diplotype and the phenotype class derived from it. How a phenotype affects a dose or drug choice is not the caller's task but that of curated guidelines. CPIC and DPWG link phenotypes to concrete, evidence-based recommendations, which PharmGKB collects and rates. Genome provides this annotation through PharmCAT but keeps it clearly separate from the measurement itself; the recommendations are decision support for professionals, not an automatic therapy decision.