A Tridimensional Causal Framework for Alzheimer's Disease
Validating Genetic Risk, Molecular Manifestations, and Structural Consequences using auditable multimodal evidence infrastructure
Authors: Manuel Menéndez-González¹, Neurodiagnoses Development Team¹
¹ Fundación de Neurociencias, Spain
Abstract
Background: Current Alzheimer's disease (AD) diagnostic frameworks integrate biomarkers but lack mechanistic validation linking genetic risk to molecular manifestations and structural consequences.
Methods: We developed Neurodiagnoses, a tridimensional framework analyzing patients across three causal axes: Etiology (genetics/environment), Pathophysiology (molecular biology), and Phenotype (clinical/structural). We validated the framework using multi-cohort data (SEA-AD n=84, ADNI n=1,000, NACC n=125,000+) with standard data science methods (Python/pandas/scikit-learn).
Results: Causal validation (Axis 1→2) demonstrated that APOE ε4 carriers showed significantly higher molecular pathology across 5/8 markers (validation score 5/8): tau pathology (Cohen's d=1.00, p<0.001), amyloid plaques (d=1.35, p<0.001), phospho-tau (d=0.58, p=0.018), Aβ40 (d=0.72, p=0.003), and astrogliosis (d=0.54, p=0.026). Spatial validation (Axis 2→3) revealed 12 significant correlations between regional molecular pathology and structural atrophy (validation score 12/32 correlations, average |r|=0.406).
Conclusions: The tridimensional framework successfully validates mechanistic links from genetic risk through molecular pathology to structural brain changes, providing causal evidence beyond current biomarker-based approaches. Transparent implementation using standard tools demonstrates this is not a "black box" but follows conventional biostatistical methods.
Key Findings
Causal Validation Score
APOE ε4 genetic risk actively manifests at molecular level. 5/8 markers validated with large effect sizes (Cohen's d up to 1.35).
Spatial Validation Score
Molecular pathology correlates with regional atrophy. 12/32 region-pathology links validated (r=-0.35 to -0.60).
⚠️ Note on Reproducibility: The manuscript presents TWO complementary analyses: (1) Direct statistical validation using scripts (causal validation score 5/8) - maximum scientific rigor, (2) Web-based system demonstration using APIs (variable scores) - web platform implementation. Both are valid but serve different purposes. See WEB_REPLICATION_GUIDE.md for details.
Transparency & Reproducibility
🔍 Not a Black Box
All analyses use standard, open-source Python libraries: pandas, scipy, scikit-learn, matplotlib. Every statistical test is documented with code examples in the manuscript.
📊 Standard Methods
Causal validation: Independent t-tests, Cohen's d effect sizes. Spatial validation: Pearson correlations. No proprietary algorithms - Statistics 101.
💻 Code Available
Complete analysis scripts provided. Any researcher with basic Python knowledge can reproduce results.