Integrated Quantitative Systems Pharmacology model of Cold Agglutinin Disease, linking clonal marrow B-cell/lymphoplasmacytic monoclonal IgM cold-agglutinin production to cold-dependent RBC binding, classical complement pathway activation (C1q-C1r-C1s → C4 → C3), C3b-mediated extravascular hepatic Kupffer-cell hemolysis, C5b-9 (MAC)-mediated intravascular hemolysis, and downstream hemolysis biomarkers (Hb, reticulocytes, bilirubin, LDH, haptoglobin) — with the modern pharmacology stack (rituximab, bendamustine, sutimlimab, pegcetacoplan) and the clinical endpoints used to track and manage this steroid-refractory disease.

Cold Agglutinin Disease (CAD) is a rare, clonal B-cell/lymphoplasmacytic lymphoproliferative disorder of the bone marrow (IGHV4-34-restricted, typically MYD88 wild-type — distinguishing it from Waldenström macroglobulinemia) in which the malignant clone secretes a monoclonal IgM with cold-dependent anti-I/i specificity. At peripheral/acral temperatures below core body temperature, this IgM binds RBC surface antigens and agglutinates red cells, and its pentameric structure is a potent activator of the classical complement pathway: C1q binding drives C1r/C1s serine protease activation, C4 and C3 cleavage, and C3b deposition on the RBC surface. Opsonized cells are predominantly cleared by hepatic Kupffer-cell CR1/CRIg-mediated erythrophagocytosis (extravascular hemolysis), while a variable fraction proceeds to full C5b-9 membrane attack complex (MAC) formation and intravascular hemolysis. The result is chronic hemolytic anemia with reticulocytosis, indirect hyperbilirubinemia, elevated LDH, and haptoglobin depletion, punctuated by acute cold-exposure hemolytic crises and accompanied by acrocyanosis/Raynaud-like phenomena. Critically, CAD is characteristically steroid-refractory (unlike warm AIHA), so therapy targets either the clonal B-cell compartment (rituximab, alone or with bendamustine) or the complement cascade directly (sutimlimab, an anti-C1s antibody approved for CAD; pegcetacoplan, an investigational C3 inhibitor with dual classical+alternative pathway blockade).

1. Genetic / clonal predisposition (MYD88 wild-type, KMT2D mutation, CXCR4 variant, trisomy 3, age-related immunosenescence, Mycoplasma/EBV infection triggers for secondary CAD)
2. Clonal B-cell lymphoproliferation & IgM production (bone marrow clonal B-cell population, IL-6/BAFF microenvironment, plasmacytoid differentiation, IgM pentamer assembly, anti-I/anti-i specificity, free light chain)
3. Cold-dependent antibody-RBC binding (peripheral/acral cooling, thermal amplitude, IgM conformational change, RBC I/i antigen, agglutination, microvascular sludging, rewarming dissociation)
4. Classical complement pathway activation (C1q-C1r-C1s, C4/C2 cleavage, C3 convertase C4b2a, C3b/iC3b/C3d deposition, RBC complement regulators CD55/CD59/MCP)
5. Terminal complement / MAC (C5 convertase, C5a anaphylatoxin, C5b-6-7 complex, C5b-9 MAC, RBC membrane pore, intravascular hemolysis, free plasma hemoglobin)
6. Extravascular hemolysis — liver Kupffer cells (CR1/CRIg receptor, phagocytosis, erythrophagocytosis, partial spherocyte formation, hepatic iron recycling)
7. Hemolysis consequences / biomarkers (hemoglobin, anemia severity, reticulocytes, indirect bilirubin, LDH, haptoglobin depletion, hemoglobinuria, compensatory erythropoiesis/EPO)
8. Clinical manifestations (fatigue, acrocyanosis, Raynaud-like phenomenon, livedo reticularis, cold-induced hemolytic crisis, jaundice, exertional dyspnea, cognitive fatigue, QoL/FACIT-fatigue)
9. Associated/secondary conditions (lymphoplasmacytic lymphoma transformation, Waldenström overlap, secondary cold agglutinin syndrome, complement-mediated thrombosis risk)
10. Drug PK/PD — rituximab (anti-CD20 B-cell depletion, ADCC/CDC lysis, IgM titer decline, relapse via B-cell repopulation)
11. Drug PK/PD — bendamustine (DNA cross-linking, clonal apoptosis, marrow suppression, synergy with rituximab)
12. Drug PK/PD — sutimlimab (anti-C1s target binding/TMDD, classical-pathway blockade, rapid Hb response, infusion-reaction risk)
13. Drug PK/PD — pegcetacoplan (C3/C3b binding, C3-convertase blockade, dual classical+alternative pathway inhibition, breakthrough-hemolysis risk)
14. Clinical endpoints & management (transfusion requirement, Hb-response criteria, steroid-refractory status, cold avoidance, folate supplementation, thromboprophylaxis)

• Drug PK (8 compartments) — rituximab (central + peripheral), bendamustine (central + peripheral), sutimlimab (central + peripheral), pegcetacoplan (SC depot + central). As with other surrogates in this library, each drug compartment is a dose-proportional "exposure" state mapped through an Emax function onto its mechanistic target, not a literal multi-compartment plasma PK/TMDD model (though sutimlimab's saturable elimination is approximated).
• Clonal / complement axis (7 compartments) — clonal B-cell burden, monoclonal IgM titer, an acute cold-exposure pulse driver, active C1s enzymatic state, C4, C3, C3b RBC-surface opsonization, and C5b-9 MAC (8 compartments total including the cold-pulse driver).
• Hemolysis biomarkers (5 compartments) — hemoglobin, reticulocyte percentage, indirect bilirubin, LDH, haptoglobin.

1. Patient profile — severity slider, secondary-CAD flag, presenting Hb, ambient cold exposure, cold-avoidance counseling flag.
2. Drug PK — fractional drug-effect trajectories for all four agents.
3. Complement pathway PD — classical-pathway flux/C4/C3, C3b opsonization & MAC, clonal burden & IgM titer.
4. Hemolysis biomarkers — hemoglobin & reticulocytes, bilirubin & LDH, haptoglobin.
5. Clinical endpoints — endpoint summary table (Day-0, Day-30, Day-90, Day-end).
6. Scenario comparison — runs all 7 built-in scenarios with the chosen patient profile.
7. Cold-exposure challenge — dedicated view of the acute cold-pulse scenario's transient MAC/hemolysis spike and Hb dip.
8. References — links to the full bibliography.

# 1) Render the mechanistic map
dot -Tsvg cad_qsp_model.dot -o cad_qsp_model.svg
dot -Tpng -Gdpi=150 cad_qsp_model.dot -o cad_qsp_model.png

# 2) Simulate scenarios in R
install.packages(c("mrgsolve","dplyr","tidyr","ggplot2","shiny",
                   "shinydashboard","DT"))
library(mrgsolve)
mod <- mread_cache("cad_mrgsolve_model.R")
out <- mrgsim(mod, ev(amt=1.3, cmt="SUT_CENT", ii=7, addl=1) +
                    ev(amt=1.1, cmt="SUT_CENT", time=14, ii=14, addl=11), end=180)
plot(out, c("Hemoglobin","MAC_level","Clonal_burden","LDH_level"))

# 3) Launch the dashboard
shiny::runApp("cad_shiny_app.R")

• Designed for research, education, and hypothesis generation — not clinical decision support.
• Drug "exposure" compartments are pragmatic dose-proportional surrogates for sustained mechanistic exposure, not literal multi-compartment plasma PK/TMDD models.
• Complement compartments (C1s, C4, C3, C3b, MAC) are simplified lumped states representing pathway flux, not a full biophysical cascade model with individual rate constants for every convertase step.
• The extravascular-vs-intravascular hemolysis split (W_C3B_EV vs W_MAC_IV) is a qualitative weighting consistent with CAD's known predominance of extravascular (Kupffer-cell) clearance, not a validated quantitative partition.
• Corticosteroids are intentionally not modeled as a treatment arm, reflecting CAD's well-documented steroid-refractory nature — this is a deliberate omission, not an oversight.

Inherits the repository LICENSE.