Causal Precision Pharmacotherapy from Genotype to Dose through Phenoconversion, Treatment History, Time-Varying Clinical Context, and Competing Therapeutic Goals

Pharmacogenomics can identify inherited determinants of drug disposition, response, and toxicity, but genotype alone cannot specify an optimal treatment for a patient whose functional phenotype, disease state, concomitant therapy, prior exposure, therapeutic priorities, and available options change over time. This Original Causal Decision Architecture Article proposes a Causal Precision Pharmacotherapy Decision Architecture that separates germline information from baseline functional phenotype, dynamically updates phenotype through phenoconversion, represents treatment history and current clinical context as a time-indexed patient state, and compares feasible drug-dose-monitoring strategies through counterfactual outcome profiles. The architecture treats efficacy, toxicity, feasibility, treatment burden, and equity as competing rather than interchangeable objectives. It introduces an explicit uncertainty layer through which missing decision-critical information, weak transportability, causal ambiguity, model miscalibration, or unresolved trade-offs may trigger additional measurement, intensified monitoring, restricted use, or abstention. Counterfactual dose and therapy comparison is framed as an estimand-specific process rather than selection of the option with the most favorable prognostic prediction. Validation is distributed across genotype interpretation, dynamic phenotype updating, causal identification, model verification, calibration, temporal and external transportability, clinical utility, human-system interaction, and governance. The original contribution is the integration of these elements into a non-substitutable architecture in which no genomic, pharmacokinetic, computational, or clinical component is sufficient alone. The proposal requires computational, experimental, clinical, implementation, and equity-focused validation. It does not establish a validated predictive model, universal treatment rule, clinical recommendation, regulatory pathway, or deployment-ready pharmacogenomic digital twin.