알츠하이머병(Alzheimer's Disease, AD)은 가장 흔한 신경퇴행성 질환으로, 전 세계적으로 약 5,500만 명, 한국에서는 약 90만 명이 이 질환을 앓고 있습니다. AD는 서서히 진행하는 인지기능 저하를 특징으로 하며, 기억력 손상에서 시작하여 언어, 판단, 일상생활 능력의 전반적인 저하로 이어집니다.
아밀로이드 캐스케이드 가설 (Amyloid Cascade Hypothesis)
AD의 주요 발병기전으로 널리 받아들여지는 아밀로이드 캐스케이드 가설에 따르면, 아밀로이드 전구 단백질(APP)의 비정상적 처리로 인해 생성된 아밀로이드 베타(Aβ42) 펩타이드가 뇌 조직에 축적되어 노인반(senile plaque)을 형성합니다. 이러한 Aβ42 축적은 신경염증 반응을 유발하고, 시냅스 기능 장애 및 신경세포 사멸로 이어집니다.
타우 병리 (Tau Pathology)
타우 단백질의 과인산화(hyperphosphorylation)는 신경섬유다발(neurofibrillary tangles, NFT)의 형성으로 이어집니다. 타우 병리는 내후각 피질에서 시작하여 해마, 신피질로 전파되며(Braak 병기), 인지기능 저하와 밀접하게 연관됩니다.
신경염증 (Neuroinflammation)
Aβ 축적에 반응하여 미세아교세포(microglia)와 별아교세포(astrocyte)가 활성화됩니다. 초기에는 Aβ 제거를 위한 보호적 반응이지만, 만성 활성화 상태에서는 친염증성 사이토카인(IL-1β, TNF-α, IL-6) 분비가 증가하여 신경세포 손상을 악화시킵니다.
콜린성 시스템 손상 (Cholinergic System Impairment)
마이네르트 기저핵(nucleus basalis of Meynert)의 콜린성 신경세포 소실은 대뇌 피질과 해마의 아세틸콜린(ACh) 농도 감소를 초래합니다. 이는 인지기능, 특히 기억력과 주의력 저하에 직접적으로 기여하며, 도네페질 등 아세틸콜린에스테라제(AChE) 억제제 치료의 근거가 됩니다.
Alzheimer's disease (AD) is the most common neurodegenerative disorder and the leading cause of dementia worldwide. It accounts for 60–80% of all dementia cases, affecting approximately 55 million people globally and approximately 900,000 people in South Korea.
Pathological Hallmarks
- Amyloid beta plaques: Extracellular deposits of aggregated Aβ42 peptides, formed by sequential cleavage of APP by β-secretase (BACE1) and γ-secretase
- Neurofibrillary tangles (NFTs): Intracellular accumulations of hyperphosphorylated tau protein, causing microtubule destabilization and neuronal death
- Neuroinflammation: Chronic activation of microglia and astrocytes, contributing to synaptic loss and neurodegeneration
- Vascular changes: Cerebral amyloid angiopathy (CAA), impaired blood-brain barrier (BBB), reduced cerebral blood flow
Genetic Factors
- APOE4: Major genetic risk factor for late-onset AD; increases Aβ accumulation and reduces clearance (1 allele: 3× risk; 2 alleles: 12× risk)
- PSEN1/PSEN2: Presenilin mutations cause early-onset familial AD by altering γ-secretase activity, increasing Aβ42/Aβ40 ratio
- APP mutations: Rare mutations near secretase cleavage sites cause familial AD or protective effects (A673T)
Disease Stages
| Stage | Description | Cognitive Status | Biomarker Changes |
|---|---|---|---|
| Preclinical | No symptoms | Normal | Aβ PET+, CSF Aβ42↓ |
| MCI due to AD | Subtle changes | Objective impairment, independent ADL | Tau PET+, MRI atrophy |
| Mild AD | Memory/cognitive decline | Impaired ADL | All biomarkers positive |
| Moderate AD | Significant impairment | Dependent for complex ADL | Significant neurodegeneration |
| Severe AD | Profound impairment | Fully dependent | Widespread cortical atrophy |
The AD QSP model integrates five mechanistic components representing the key pathophysiological pathways and pharmacological interventions:
This component models the sequential processing of amyloid precursor protein (APP) through both amyloidogenic and non-amyloidogenic pathways:
- APP production from neurons and metabolism via α-secretase (non-amyloidogenic) or BACE1/γ-secretase (amyloidogenic)
- Aβ40 and Aβ42 monomer production, oligomerization, and fibrillization
- Amyloid plaque formation, seeding, and growth kinetics
- Enzymatic clearance via neprilysin (NEP) and insulin-degrading enzyme (IDE)
- BBB-mediated transport (LRP1 efflux, RAGE influx)
- APOE4-dependent modulation of Aβ clearance efficiency
Models the progression of tau pathology from normal function to NFT formation:
- Normal tau synthesis and microtubule binding
- GSK3β and CDK5-mediated phosphorylation
- Hyperphosphorylated tau accumulation and oligomer formation
- NFT assembly kinetics (Braak-stage-based spread approximation)
- Tau clearance via autophagy-lysosomal pathway and proteasome
- Cross-talk with Aβ pathology (Aβ-induced GSK3β activation)
Captures the dual role of neuroinflammation in AD:
- Resting microglia activation by Aβ oligomers and DAMPs
- M1 (pro-inflammatory) vs. M2 (anti-inflammatory/phagocytic) microglial states
- Pro-inflammatory cytokine production (IL-1β, TNF-α, IL-6, IL-18)
- Astrocyte reactivity and GFAP expression
- Complement system activation (C1q, C3, CR3)
- TREM2-mediated microglial phagocytosis of Aβ
- NF-κB signaling pathway
Models the cholinergic hypothesis and synaptic function:
- ACh synthesis (ChAT activity), release, and hydrolysis (AChEI target)
- Muscarinic (M1, M2) and nicotinic (α7) receptor signaling
- Glutamatergic transmission: AMPA and NMDA receptor-mediated LTP/LTD
- NMDA receptor overstimulation and excitotoxicity (memantine target)
- Synaptic density dynamics (formation vs. Aβ/tau-mediated loss)
- BDNF-TrkB neuroprotective signaling
- Cognitive function scoring linked to synaptic integrity
Pharmacokinetic and pharmacodynamic models for four approved/emerging therapies:
Donepezil (AChE Inhibitor)
- 2-compartment oral PK model (ka, Vd, CL, brain Kp)
- Reversible AChE inhibition (IC50-based Emax model)
- ACh concentration-dependent cognitive improvement
Memantine (NMDA Receptor Antagonist)
- 3-compartment oral PK model
- Voltage-dependent, uncompetitive NMDA receptor block
- Neuroprotection via glutamate excitotoxicity attenuation
Lecanemab (Anti-Aβ Protofibril mAb)
- 2-compartment IV PK model with target-mediated drug disposition (TMDD)
- BBB penetration via IgG transcytosis (Kp ~ 0.001)
- Dose-dependent Aβ protofibrils and plaque removal
- ARIA (amyloid-related imaging abnormalities) risk modeling
Donanemab (Anti-N3pE Aβ mAb)
- Similar TMDD framework targeting N-terminally truncated Aβ (N3pE)
- Potent plaque clearance kinetics (amyloid clearance in ~76% at 76 weeks)
- Dose-suspension upon amyloid clearance (TRAILBLAZER protocol)
| File | Description |
|---|---|
ad_qsp_model.dot |
Graphviz DOT source for mechanistic map (100+ nodes, 10+ clusters) |
ad_qsp_model.svg |
Vector format mechanistic map |
ad_qsp_model.png |
PNG format mechanistic map (150 dpi) |
ad_mrgsolve_model.R |
mrgsolve ODE model with 20+ compartments, 7 treatment scenarios |
ad_shiny_app.R |
Interactive Shiny dashboard (6 tabs) |
ad_references.md |
40+ PubMed references organized by topic |
README.md |
This file |
| Parameter | Symbol | Value | Unit | Source |
|---|---|---|---|---|
| Donepezil ka (absorption) | ka_don | 1.2 | h⁻¹ | Tiseo et al. 1998 |
| Donepezil Vd | Vd_don | 0.7 | L/kg | Tiseo et al. 1998 |
| Donepezil CL | CL_don | 0.13 | L/h/kg | Tiseo et al. 1998 |
| Donepezil brain Kp | Kp_don | 0.6 | — | Estimated from tissue distribution |
| Donepezil AChE IC50 | IC50_ache | 0.006 | μg/mL | Sugimoto et al. 1995 |
| Memantine Vd | Vd_mem | 9 | L/kg | Wesemann et al. 1983 |
| Memantine CL | CL_mem | 0.13 | L/h/kg | Wesemann et al. 1983 |
| Memantine NMDAR IC50 | IC50_nmda | 1.0 | μg/mL | Parsons et al. 2007 |
| Lecanemab Vd (central) | Vd_lec | 0.05 | L/kg | Swanson et al. 2021 |
| Lecanemab t½ (terminal) | t_half_lec | 270 | h (~11 days) | Swanson et al. 2021 |
| Lecanemab CNS Kp | Kp_lec | 0.001 | — | Estimated (IgG BBB penetration) |
| Abeta42 production rate | k_prod | 0.15 | nM/h | Bateman et al. 2006 |
| Abeta42 clearance (NEP/IDE) | k_clear | 0.3 | h⁻¹ | Leissring et al. 2003 |
| Tau phosphorylation rate | k_phos | 0.05 | h⁻¹ | Estimated |
| NFT formation rate | k_agg | 0.01 | h⁻¹ | Estimated |
| Microglial activation rate | k_micro | 0.02 | h⁻¹ | Estimated |
| Synaptic damage rate | k_syn_dam | 0.03 | h⁻¹ | Estimated |
| APOE4 effect on clearance | APOE4_mult | 1.3 | fold | Liu et al. 2013 |
| Scenario | Drug(s) | Dose | Regimen | Clinical Trial Basis |
|---|---|---|---|---|
| No treatment | — | — | — | Natural history |
| Donepezil | AChE inhibitor | 10 mg | QD oral | Rogers et al. 1996 |
| Memantine | NMDAR antagonist | 20 mg | QD oral | Reisberg et al. 2003 |
| Combination | Don + Mem | 10 + 20 mg | QD oral | Tariot et al. 2004 |
| Lecanemab | Anti-Aβ mAb | 10 mg/kg | Q2W IV | CLARITY-AD (van Dyck 2023) |
| Donanemab | Anti-Aβ mAb (N3pE) | 1500 mg | Q4W IV | TRAILBLAZER-ALZ-2 (Sims 2023) |
| Triple therapy | Don + Mem + Lec | Standard | Combined | Theoretical combination |
- Citation: Van Dyck CH et al. N Engl J Med 2023;388:9-21 (PMID: 36449413)
- Design: Phase 3, randomized, double-blind, placebo-controlled
- Population: 1,795 patients with early AD (MCI or mild dementia with confirmed Aβ pathology)
- Intervention: Lecanemab 10 mg/kg IV every 2 weeks for 18 months
- Primary endpoint: Change from baseline in CDR-SB at 18 months
- Result: 27% slowing of decline vs placebo (CDR-SB change: 1.21 vs 1.66; difference -0.45, p < 0.001)
- Amyloid reduction: -55.48 centiloids vs +3.64 for placebo (p < 0.001)
- Secondary endpoints: ADAS-Cog14, ADCOMS, ADCS-MCI-ADL all significantly improved
- ARIA: ARIA-E in 12.6%, ARIA-H in 17.3% of lecanemab group
- Citation: Sims JR et al. JAMA 2023;330(6):512-527 (PMID: 37459141)
- Design: Phase 3, randomized, double-blind, placebo-controlled
- Population: 1,736 patients with early symptomatic AD (low/medium tau stratum primary analysis)
- Intervention: Donanemab 1,500 mg IV every 4 weeks (transitioned to placebo after amyloid clearance)
- Primary endpoint: Change from baseline in iADRS at 76 weeks
- Result: 35% slowing of clinical progression in combined population; 40% in low/medium tau subgroup
- Amyloid clearance: Complete clearance (< 24.1 centiloids) in 76% of patients by 76 weeks
- iADRS result: -6.02 vs -9.27 for placebo (difference 3.25, p < 0.001)
- ARIA: ARIA-E in 24.0%, ARIA-H in 31.4% of donanemab group
| Trial | Drug | N | Duration | Primary Result |
|---|---|---|---|---|
| Rogers et al. 1996 | Donepezil 5/10 mg | 468 | 24 weeks | ADAS-Cog: -2.49 (10 mg) vs +0.52 placebo |
| Reisberg et al. 2003 | Memantine 20 mg | 252 | 28 weeks | CIBIC-plus significant improvement |
| Tariot et al. 2004 | Donepezil + Memantine | 404 | 24 weeks | SIB: +0.9 vs -2.5 placebo (p = 0.002) |
# Generate SVG
dot -Tsvg ad_qsp_model.dot -o ad_qsp_model.svg
# Generate PNG at 150 dpi
dot -Tpng -Gdpi=150 ad_qsp_model.dot -o ad_qsp_model.pnglibrary(mrgsolve)
source("ad_mrgsolve_model.R")
# Runs automatically when sourced, generates simulation plots
# for all 7 treatment scenarios over 104 weekslibrary(shiny)
runApp("ad_shiny_app.R")
# Opens interactive dashboard at localhost with 6 tabs:
# 1. Patient Profile
# 2. Drug PK
# 3. PD Biomarkers
# 4. Clinical Endpoints
# 5. Scenario Comparison
# 6. Biomarker Panel| Tab | Content |
|---|---|
| 1. Patient Profile | Age, APOE4 status, baseline cognitive scores, disease stage selection |
| 2. Drug PK | Plasma and CSF concentration-time profiles for all drugs |
| 3. PD Biomarkers | Aβ42, tau, phospho-tau, neuroinflammation markers over time |
| 4. Clinical Endpoints | CDR-SB, ADAS-Cog, MMSE trajectories with clinical trial benchmarks |
| 5. Scenario Comparison | Side-by-side comparison of all 7 treatment scenarios |
| 6. Biomarker Panel | PET amyloid centiloids, CSF biomarker ratios, synaptic density |
- Simplified disease pathology: Three main pathways modeled rather than full mechanistic detail; spatial spreading (Braak staging) not explicitly implemented
- Donepezil bioavailability: Assumed 100% (actual approximately 100% based on clinical studies; well-supported assumption)
- Lecanemab CNS penetration: Modeled as passive diffusion with fixed Kp = 0.001; actual mechanism involves FcRn-mediated transcytosis and is concentration-dependent
- Tau propagation: Prion-like tau spreading not modeled spatially; Braak stage progression approximated as a monotonic progression variable
- Individual variability: Modeled as fixed effects only; no population PK/PD variability or Monte Carlo simulation implemented in base model
- Neurodegeneration irreversibility: Neuronal loss modeled as a slow, irreversible process; no neuroregeneration or plasticity mechanisms included
- ARIA mechanistic model: ARIA risk modeled empirically based on Aβ efflux rate; the true immunological cascade mediating ARIA is not fully captured
- Drug interactions: Donepezil-memantine combination modeled as additive; potential pharmacokinetic interactions (CYP2D6, CYP3A4) not explicitly simulated
- Comorbidities: Vascular contributions to cognitive impairment (VCID), cerebrovascular disease not included
- Biomarker calibration: PET amyloid centiloid scale conversion is approximate; CSF biomarker thresholds based on published reference ranges
Key references — see ad_references.md for the complete annotated list of 40+ citations:
- Jack CR Jr et al. NIA-AA Research Framework 2018 (PMID: 29653606)
- Hardy J, Selkoe DJ. The amyloid hypothesis of AD. Science 2002 (PMID: 12130773)
- Braak H, Braak E. Neuropathological stageing of Alzheimer-related changes. Acta Neuropathol 1991 (PMID: 1759558)
- Rogers SL et al. Donepezil Phase III trial. Neurology 1996 (PMID: 8649628)
- Reisberg B et al. Memantine in moderate-to-severe AD. NEJM 2003 (PMID: 12672860)
- Tariot PN et al. Memantine + donepezil combination. JAMA 2004 (PMID: 15010441)
- van Dyck CH et al. CLARITY-AD Trial 2023 (PMID: 36449413)
- Sims JR et al. TRAILBLAZER-ALZ 2 2023 (PMID: 37459141)
- Tiseo PJ et al. Donepezil PK. Br J Clin Pharmacol 1998 (PMID: 9723830)
- Swanson CJ et al. Lecanemab PK/PD. Alzheimers Res Ther 2021 (PMID: 34118953)
- Geerts H et al. QSP model for AD. CPT Pharmacometrics Syst Pharmacol 2013 (PMID: 23475765)
- Bateman RJ et al. Human amyloid beta kinetics. Nat Med 2006 (PMID: 16906156)
Model developed as part of the QSP Disease Model Library. Date: 2026-06-20.