Replicated trials and anchored models

Coupled SCLC kernel + DLL3-mediated T-cell engagement + cytokine-release-syndrome cascade. Replicates the DeLLphi-301 response rate distribution and the dose-CRS trade-off across the 10 mg and 100 mg cohorts.

Scoping: schedule rescue — can a slower step-up titration reduce grade ≥3 CRS without sacrificing response? Awaiting full appendix release for final calibration anchors.

EBV immunosurveillance ODE coupled to chronic-perturbation load (malaria prevalence). Reproduces the Moormann observation that holoendemic malaria exposure raises EBV viral load and suppresses EBV-specific T-cell surveillance — the proposed mechanistic substrate for endemic Burkitt’s geographic distribution.

Extension: mechanism-class hook — d(eBL_hazard)/d(chronic_perturbation) extends to any chronic immune perturbation, not just malaria.

41-state cardiac electrophysiology kernel — IKr, IKs, ICaL, INa, INaL and the full ionic-current set. Reproduces the published action-potential durations under reference and IKr-blockade conditions. Drives the cardiac-safety panel used in adverse-event counterfactuals.

Renin–angiotensin–aldosterone cascade coupling kidney, adrenal, and cardiac afterload. Anchor for antihypertensive titration and RAAS-targeted therapy across hypertension, HFrEF / HFpEF, and CKD.

Macrophage / foam-cell ODE coupling LDL, modified LDL, and HDL efflux with plaque-volume dynamics. Reproduces the Cohen-Myerscough baseline trajectory and the HDL-protection signal.

Extension: statin titration designed under the model — per-patient LDL trajectory + plaque progression + hepatic/myalgic side-effect constraints — with a separate identifiability map showing which parameters can be recovered from a given bloodwork panel.

Finerenone PK/PD coupled to the RAAS axis, kidney handling, and the K-gated up-titration logic the FIGARO/FIDELIO trials operationalised. Awaiting full PDF fetch for refit calibration. Direction-of-effect (eGFR preservation, CV endpoint) reproduces published anchors.

A 22-state coupled model of Ebola virus disease — viral replication, antibody response, immune activation, coagulation failure, organ failure — calibrated against all six PALM treatment arms and the observational control. Reproduces reported case-fatality rates within published confidence intervals.

Extension: the trial enrolled patients on arrival, leaving time-from-symptom to first dose as an uncontrolled confounder. The model is run with time as a free input, surfacing the 48-hour timing cliff the trial design could not quantify. The remdesivir paradox (53% CFR vs 48% observational) is shown to be consistent with allocation bias rather than drug harm.

8-state coupled HPS model — viral load, endothelial infection, capillary leak, pulmonary oedema, lactic acidosis, IL-6, platelets, bradykinin. Reproduces the Mertz 2004 null (ribavirin at median enrolment day 4.5 → ~13% mortality, no endpoint benefit) AND the Safronetz 2011 positive (ribavirin at day 1 → 0% mortality in hamster model translated to human physiology).

The wrong-stage thesis confirmed in silico. Companion piece on the blog.

Coupled immune + CNS + cardiovascular trajectory model for paediatric post-acute COVID. Replicates the 3-year cohort distribution of symptom persistence and recovery. Collaboration with the ISARIC paediatric long COVID consortium.

Coupled cardiac electrophysiology + immune signalling kernel. Compares the brief, UPS-cleared spike exposure of vaccination with the sustained, endothelium-resident spike of active infection. The hazard ratio falls out cleanly at 6–7× against the vaccine — consistent with the published surveillance literature. The kinetic story is in the model, not in counting events.

Artesunate → DHA hepatic-esterase hydrolysis kernel coupled to plasma, parasite load, and a ferroptosis-pathway hook for the Efferth-group oncology framing. Reproduces the Krishna PK series; extends to glioma and other DHA-sensitive tumour panels under bone-marrow and hepatic toxicity constraints.

Hepatitis C kernel coupled to hepatocyte infection, viral replication, and combination-DAA pharmacology. Reproduces the published SVR rates across the licensed combinations; used as a worked example of cross-company combination comparison under a shared body model.

14-state osteoblast/osteoclast/osteocyte coupling with PTH, RANKL, OPG, sclerostin, and mechanical-load levers. Reproduces Lemaire’s baseline steady state, Pivonka’s TGF-β extension, and the Graham-Ayati Wnt/sclerostin mechanical-loading response. Drives the osteoporosis-titration program and the bone-metastatic component of the prostate and breast cancer programs.

Thyroid-hormone-receptor-β agonist coupled to hepatic lipid handling, NASH inflammation, and fibrosis trajectory. Reproduces the MAESTRO-NASH primary endpoint distributions; sweep against alternative dosing schedules in progress.

Composed CNS pharmacology + blood-brain-barrier transport + anti-amyloid antibody binding + microglial activation. Reproduces both pivotal trials’ 18-month cognitive endpoints and the reported ~0.5% brain-volume reduction within confidence intervals.

Extension: the model is run past month 30 under both the pseudoatrophy and the real-neurotoxicity hypotheses. The two predictions diverge at the post-trial horizon — the published 18-month data is consistent with both. Anticoagulation + ApoE4 homozygote sub-population simulated separately.

Reproduces Friberg’s transit-chain model (proliferating cells → transit → circulating neutrophils) with drug-effect on proliferation. Used as the marrow-toxicity constraint for every chemotherapy schedule search in the dose-and-schedule program.

The virtual-cohort generation method used across every trial-replication program. Sample patient covariate vectors from the population distribution, simulate the trial protocol, check that endpoint distributions land within published confidence intervals, then re-sample for the counterfactual cohort. The infrastructure layer beneath every other entry on this page.