[M5][PKPD] Implement multi-drug PK/PD, combination, and treatment-sequence simulation

[andreazedda]

Objective

Extend the single-drug and simplified schedule logic into a versioned multi-drug PK/PD framework capable of representing combinations, cycles, interruptions, sequence effects, organ constraints, and resistance-linked response.

Initial therapy classes

• IMiDs;
• proteasome inhibitors;
• anti-CD38 antibodies;
• corticosteroids;
• alkylating agents and transplant conditioning abstractions;
• supportive interventions where they materially affect observations or toxicity.

Scope

• drug-specific dose, route, schedule, PK parameters, exposure, and uncertainty;
• simultaneous and sequential combinations;
• interaction models classified as additive, synergistic, antagonistic, or unknown;
• time-varying clone sensitivity and organ-state constraints;
• treatment hold, restart, dose reduction, and recovery windows;
• explicit separation of PK, PD, disease response, and toxicity;
• evidence-linked parameter registry;
• model reduction when parameters are unavailable.

Acceptance criteria

• Each drug has an independent versioned parameter and evidence record.
• Combination effects are not assumed synergistic without a defined model.
• Sequence and schedule changes can alter outputs even when cumulative dose is equal.
• Organ and marrow states can constrain exposure or toxicity.
• Interaction uncertainty is propagated.
• Zero exposure produces no direct drug effect.
• Solver convergence, dimensional consistency, and numerical regression tests pass.
• The report distinguishes literature-derived, estimated, and heuristic interactions.

Non-goals

This framework compares mechanistic scenarios; it does not prescribe a regimen.