ALZHEIMER'S ASSOCIATION AAIC 26

Coherence-Validated Causal World Models for Multi-Scale Alzheimer's Disease Progression and Pharmacologic Reversal

David Scott Lewis, Enrique Zueco, Haley Yi
AIXC Bio, Zaragoza, Spain | reports@aixcbio.com

KEY TAKEAWAY: C3WM guarantees biologically sound in silico simulation of AD treatment by validating causal and temporal coherence across scales.

INTRODUCTION
AD progression across multi-scales
MOLECULAR DISRUPTIONS (Energy & Redox Homeostasis) -> CELLULAR STRESS -> NEUROINFLAMMATION -> NEUROVASCULAR DYSFUNCTION -> COGNITIVE DECLINE

Breakthrough: physiological brain NAD+ homeostasis restoration (e.g., with P7C3-A20) can reverse advanced pathology and behavioral deficits in symptomatic mouse models.

Critical Need: for action-conditioned AI to retain authentic causal connections and comply with biological couplings, translating mouse findings to human patients.

C3WM Data Integration Core
Molecular Data (Mass Spec) -> C3WM Model Integration -> System-Scale Data (fMRI)
Cellular Data (Confocal Microscopy) -> C3WM Model Integration -> Cognitive Data (Test Scores)
AIXC Bio

METHODS - C3WM Framework
Hierarchical, action-conditioned latent dynamics model

AUDITABLE CAUSAL SCAFFOLD over interpretable disease modules
Key MOLECULAR pathways -> NEUROVASCULAR pathways -> Biological links
MOLECULAR pathways -> NEUROINFLAMMATION -> NEURAL pathways

Monte Carlo Wavelet Coherence (MCWC) VALIDATION PROTOCOL
Coherence metrics act as training regularizers -> provide conservative uncertainty estimation

Illustrative example, correct temporal delays:
NAD+ restoration -> reduced neuroinflammation -> recovery of synaptic plasticity

Standard Deep Learning Models:
Good MSE, bad simulator

C3WM:
Good MSE, good simulator
Improved calibration with coherence auditing
Reduced out-of-distribution errors and localized failure correction

RESULTS
Evaluation against the endpoints
REVERSAL-CENTRIC BENCHMARK
- NAD+/NADH ratios
- Proteomic Signatures
- Oxidative Damage
- Blood-Brain Barrier Integrity
- Synaptic Plasticity
- Behavior

CONCLUSIONS
- Formally validated, multi-scale in silico environment for AD simulation
- Ensures AI-generated treatment trajectories preserve true causal and temporal biological couplings
- Highly reliable, auditable tool for testing novel interventions and optimizing complex dosing regimens before in vivo deployment.

Roadmap:
AI simulation -> Optimized dosing strategies
AIXC Bio