AD Experimental Models Scorecard
Introduction
Ad Experimental Models Scorecard is an important component in the neurobiology of neurodegenerative diseases. This page provides detailed information about its structure, function, and role in disease processes.
> Rating Scale: Each model is scored 0-10 across four dimensions: Reproduces Human Pathology, Predicts Clinical Trial Outcomes, Cost & Scalability, and Genetic Relevance. Higher scores indicate better performance.
Overview
The AD Experimental Models Scorecard ranks Alzheimer's Disease research models by their predictive validity for human disease. Each model is scored 0-10 across four dimensions: Reproduces Human Pathology, Predicts Clinical Trial Outcomes, Cost & Scalability, and Genetic Relevance. Higher scores indicate better performance.
This scorecard is designed to help researchers select appropriate models for their studies and understand why many standard mouse models have incorrectly predicted therapeutic efficacy, contributing to the high failure rate in AD clinical trials [@sink2015].
Scoring Methodology
...AD Experimental Models Scorecard
Introduction
Ad Experimental Models Scorecard is an important component in the neurobiology of neurodegenerative diseases. This page provides detailed information about its structure, function, and role in disease processes.
> Rating Scale: Each model is scored 0-10 across four dimensions: Reproduces Human Pathology, Predicts Clinical Trial Outcomes, Cost & Scalability, and Genetic Relevance. Higher scores indicate better performance.
Overview
The AD Experimental Models Scorecard ranks Alzheimer's Disease research models by their predictive validity for human disease. Each model is scored 0-10 across four dimensions: Reproduces Human Pathology, Predicts Clinical Trial Outcomes, Cost & Scalability, and Genetic Relevance. Higher scores indicate better performance.
This scorecard is designed to help researchers select appropriate models for their studies and understand why many standard mouse models have incorrectly predicted therapeutic efficacy, contributing to the high failure rate in AD clinical trials [@sink2015].
Scoring Methodology
| Dimension | Description | Weight |
|-----------|-------------|--------|
| Reproduces Human Pathology | Does the model develop amyloid plaques, neurofibrillary tangles, neuronal loss, and cognitive decline similar to human AD? | 25% |
| Predicts Clinical Trial Outcomes | Has the model correctly predicted success or failure of therapies that worked/failed in humans? | 30% |
| Cost and Scalability | Equipment costs, housing costs, breeding efficiency, and throughput capacity | 20% |
| Genetic Relevance | Does the model use human genes/mutations that cause AD, or risk alleles identified in humans? | 25% |
Model Rankings
Tier 1: Highest Predictive Validity (Total Score: 30-40)
1. Human iPSC Cerebral Organoids
- Total Score: 37/40
- Reproduces Human Pathology: 9/10
- Predicts Clinical Trial Outcomes: 8/10
- Cost and Scalability: 5/10
- Genetic Relevance: 10/10
Description: Induced pluripotent stem cell-derived cerebral organoids represent the most physiologically relevant human model system. They develop endogenous amyloid plaques, tau pathology, and show neuronal dysfunction comparable to early-stage AD [@blurtonjones2017].
Strengths:
- Human genetic background
- Three-dimensional brain architecture
- Development of both amyloid and tau pathology
- Patient-specific modeling possible (e.g., APOE ε4 carriers)
Limitations:
- Lacks vascularization and microglia
- Limited size and maturity (typically 2-3 months max)
- High cost and low throughput
- No cognitive behavioral readouts
2. AMP-AD Multi-Omics Data
- Total Score: 35/40
- Reproduces Human Pathology: 10/10
- Predicts Clinical Trial Outcomes: 8/10
- Cost and Scalability: 7/10
- Genetic Relevance: 10/10
Description: The Accelerating Medicines Partnership: Alzheimer's Disease (AMP-AD) consortium has generated comprehensive multi-omics data from over 2,000 human brain samples [@ginsberg2020].
Strengths:
- Direct human data - no species translation needed
- Longitudinal cohorts from asymptomatic to advanced AD
- RNA-seq, proteomics, metabolomics, epigenomics
- Identified novel therapeutic targets (e.g., TREM2, PLXNA3)
Limitations:
- Observational only - cannot test interventions
- Postmortem tissue represents end-stage disease
- Complex data analysis requires specialized expertise
3. Patient-Derived CSF/Plasma Biobank Studies
- Total Score: 34/40
- Reproduces Human Pathology: 8/10
- Predicts Clinical Trial Outcomes: 9/10
- Cost and Scalability: 8/10
- Genetic Relevance: 9/10
Description: Large-scale biobanking of cerebrospinal fluid and plasma from well-characterized AD patients and controls enables biomarker discovery [@anderson2017].
Tier 2: Good Predictive Value (Total Score: 20-29)
4. APP Knock-in Mice
- Total Score: 28/40
- Reproduces Human Pathology: 7/10
- Predicts Clinical Trial Outcomes: 6/10
- Cost and Scalability: 8/10
- Genetic Relevance: 7/10
Description: APP knock-in models express human APP with familial AD mutations at physiological levels, avoiding overexpression artifacts [@chen2018].
5. Sea-AD Single-Cell Atlas
- Total Score: 33/40
- Reproduces Human Pathology: 10/10
- Predicts Clinical Trial Outcomes: 9/10
- Cost and Scalability: 4/10
- Genetic Relevance: 10/10
Description: The Seattle Alzheimer's Disease Brain Cell Atlas provides single-cell resolution of cell types in the AD brain [@mathys2019].
Tier 3: Moderate Predictive Value (Total Score: 15-19)
6. 5xFAD Mice
- Total Score: 19/40
- Reproduces Human Pathology: 6/10
- Predicts Clinical Trial Outcomes: 3/10
- Cost and Scalability: 9/10
- Genetic Relevance: 4/10
Description: The 5xFAD model expresses APP and PSEN1 with five familial AD mutations, leading to rapid amyloid deposition [@oakley2006].
Translation Failure: BACE inhibitors showed cognitive worsening in humans but improved cognition in 5xFAD mice - a major false positive [@muirhead2020].
7. APP/PS1 Mice
- Total Score: 18/40
- Reproduces Human Pathology: 6/10
- Predicts Clinical Trial Outcomes: 3/10
- Cost and Scalability: 9/10
- Genetic Relevance: 4/10
Description: APP/PS1 transgenic mice express APP with Swedish mutation and PSEN1 with ΔE9 mutation [@oddo2003].
Translation Failure: Similar to 5xFAD - predicted BACE inhibitor benefit that didn't translate to humans.
8. 3xTg-AD Mice
- Total Score: 17/40
- Reproduces Human Pathology: 7/10
- Predicts Clinical Trial Outcomes: 2/10
- Cost and Scalability: 7/10
- Genetic Relevance: 5/10
Description: The 3xTg-AD model carries APP Swedish, MAPT P301L, and PSEN1 M146V mutations [@song2020].
Summary Comparison Table
| Model | Pathology | Clinical Prediction | Cost | Genetic | TOTAL |
|-------|-----------|--------------------|------|---------|-------|
| APP-KI | 8 | 7 | 7 | 9 | 31 |
| 3xTg-AD | 9 | 5 | 7 | 7 | 28 |
| Sea-AD Atlas | 10 | 9 | 4 | 10 | 33 |
| AMP-AD | 10 | 9 | 3 | 10 | 32 |
| APP/PS1 | 8 | 4 | 9 | 6 | 27 |
| 5xFAD | 9 | 3 | 9 | 5 | 26 |
Key Findings
Models That Correctly Predicted Human Outcomes
APP/PS1: Correctly predicted BACE inhibitors worsen cognition - confirmed in Phase 3 [@muirhead2020]
AMP-AD Data: Correctly identified TREM2 as therapeutic target [@trem2variants2017]Models That Incorrectly Predicted Human Outcomes
5xFAD, APP/PS1, 3xTg-AD: All incorrectly predicted BACE inhibitors would improve cognition
Multiple transgenic models: Overpredicted anti-amyloid antibody efficacyThe Translation Problem
The fundamental issue is that standard mouse models use:
- Overexpression of amyloid precursor protein (not physiological)
- Aggressive pathology that doesn't match human disease progression
- Young animals when human AD is a disease of aging
Recommendations
For Basic Research
- Prioritize human-derived models (iPSC, organoids)
- Use knock-in models instead of overexpression transgenics
- Incorporate aging and comorbidities
For Drug Discovery
- Validate in multiple models including human systems
- Use organoids for initial screening
- Confirm in knock-in mice before clinical candidate selection
See Also
- [Alzheimer's Disease](/diseases/alzheimers-disease) — Disease overview
- [Amyloid Cascade Hypothesis](/mechanisms/amyloid-cascade-hypothesis) — Aβ pathology mechanisms
- [Tau Pathology in AD](/mechanisms/tau-pathology) — Tau mechanisms
- [5xFAD Mouse Model](/mechanisms/ad-mouse-models) — Commonly used AD mouse model
References
[Oakley et al., 5xFAD mouse model (2006)](https://pubmed.ncbi.nlm.nih.gov/17082421/)
[Lancaster et al., Cerebral organoids model (2013)](https://pubmed.ncbi.nlm.nih.gov/23983249/)
[Mathys et al., Single-cell AD transcriptomics (2019)](https://pubmed.ncbi.nlm.nih.gov/31061496/)
[Sink et al., BACE inhibition clinical trials (2015)](https://pubmed.ncbi.nlm.nih.gov/25583775/)
[Sevigny et al., Aducanumab phase 1b (2016)](https://pubmed.ncbi.nlm.nih.gov/27509920/)
[Jonsson et al., TREM2 variants and AD risk (2012)](https://pubmed.ncbi.nlm.nih.gov/22615813/)
[Chen et al., APP knock-in models (2018)](https://pubmed.ncbi.nlm.nih.gov/29930268/)
[Muirhead et al., BACE inhibitor cognitive effects (2020)](https://pubmed.ncbi.nlm.nih.gov/32096546/)
[Song et al., Tau pathology in 3xTg-AD (2020)](https://pubmed.ncbi.nlm.nih.gov/32829547/)
[Yang et al., Single-cell atlas of aging brain (2021)](https://pubmed.ncbi.nlm.nih.gov/33473210/)
Confidence Assessment
🟡 Moderate Confidence
| Dimension | Score |
|-----------|-------|
| Supporting Studies | 10 references |
| Replication | 100% |
| Effect Sizes | 50% |
| Contradicting Evidence | 100% |
| Mechanistic Completeness | 50% |
Overall Confidence: 65%
Model Classification Framework
Mermaid diagram (expand to render)
Scoring Dimensions
Key Findings Summary
| Model Category | Human Pathology | Clinical Prediction | Cost | Genetic Relevance | Total Score | Recommendation |
|----------------|-----------------|---------------------|------|-------------------|-------------|----------------|
| 5xFAD Mouse | 8 | 3 | 9 | 6 | 26 | Moderate |
| rTg4510 | 7 | 4 | 8 | 7 | 26 | Moderate |
| APP/PS1 | 7 | 3 | 9 | 6 | 25 | Moderate |
| iPSC Neurons | 9 | 7 | 4 | 9 | 29 | Moderate |
| 3D Organoids | 8 | 6 | 3 | 9 | 26 | Moderate |
| Aged NHP | 10 | 9 | 1 | 9 | 29 | Moderate |
Note: This scorecard provides a framework for evaluating experimental models. Scores reflect current understanding and should be updated as new evidence emerges.