[M5][IMMUNOTHERAPY] Implement modality-specific CAR-T, bispecific, and ADC research models

[andreazedda]

Objective

Implement a common but modality-aware framework for antibody-drug conjugates, bispecific antibodies, and engineered cellular therapies used or investigated in Multiple Myeloma.

Scope

Shared

• target and antigen-density profiles;
• on-target/off-tumour exposure;
• antigen loss or downregulation;
• tumour and normal-compartment selectivity;
• clone-specific response and escape;
• modality-specific uncertainty and evidence.

ADC

• antibody binding, internalization, linker stability, payload release, bystander effect, systemic leakage, and payload toxicity.

Bispecific antibodies

• target engagement, T-cell recruitment, exposure, activation, exhaustion, cytokine signal, infection and immune-suppression proxies.

CAR-T and related cellular therapies

• cell dose, expansion, persistence, trafficking abstraction, activation threshold, exhaustion, antigen escape, cytokine toxicity, and optional safety switches.

Acceptance criteria

• Target and modality are distinct model objects.
• BCMA, GPRC5D, FcRH5, CD38, SLAMF7, and future targets can use the same contracts without assuming identical biology.
• ADC, bispecific, and CAR-T toxicity mechanisms remain separate.
• Antigen loss and T-cell dysfunction are explicit dynamic mechanisms.
• Normal-compartment targeting is represented.
• Every coefficient has an evidence or heuristic classification.
• Modality ablations and sensitivity analyses are reported.
• Outputs remain research scenarios rather than clinical recommendations.

Non-goals

The first implementation is a mechanistic research framework and not a validated predictor of individual immunotherapy response or toxicity.