Neurodiagnoses Web Platform Achieves Complete Equivalence with Traditional Biostatistical Methods for Multi-Omics Causal Validation in Alzheimer's Disease
Running Title: Web Platform Equivalence for Causal Validation
Authors: Manuel Menéndez1, Neurodiagnoses AI Development Team1
1 Fundación de Neurociencias, Spain
Correspondence: manuelmenendez@fneurociencias.org
Abstract
Background: Advanced multi-omics causal analyses in Alzheimer's disease (AD) traditionally require programming expertise and local computational infrastructure, limiting accessibility to specialized research centers.
Methods: We developed Neurodiagnoses, a web-accessible platform for tridimensional causal validation (Genetics→Molecular→Structural), and validated its equivalence against traditional Python-based biostatistical methods. Using multi-cohort data (SEA-AD n=84, ADNI n=1,000), we compared web platform APIs against conventional scripts (scipy.stats, pandas) for causal validation (case-control t-tests) and spatial validation (Pearson correlations).
Results: Perfect equivalence achieved between web platform and traditional methods. Causal validation (Genetics→Molecular): Neurodiagnoses API 62.5% vs Traditional scripts 62.5% (0.0% deviation, 5/8 biomarkers significant). Spatial validation (Molecular→Structural): Neurodiagnoses API 37.5% vs Traditional scripts 37.5% (0.0% deviation, 12/32 correlations significant). Both methods identified identical biomarkers: tau pathology (d=1.00, p<0.001), amyloid plaques (d=1.35, p<0.001), phospho-tau (d=0.58, p=0.018). Platform calibration achieved maximum statistical validation (p<0.000001, d=1.452). Strongest spatial associations: tau×inferior temporal volume (r=-0.602, p<0.001), tau×middle temporal (r=-0.497, p=0.003) - identical across methods.
Conclusions: The Neurodiagnoses web platform achieves complete methodological equivalence with traditional biostatistical scripts (causal validation: 0.0% deviation, spatial validation: 0.0% deviation, identical biomarkers and effect sizes). This demonstrates that sophisticated multi-omics causal analyses can be democratized via web interfaces without sacrificing statistical rigor. The platform provides both research-grade cohort analyses (matching journal standards) and clinical-grade individual assessments, proving that web-based AI can deliver the best of both worlds: scientific validity with global accessibility.
Keywords: Alzheimer's disease, Web-based analysis, Methodological equivalence, Multi-omics, Causal validation, APOE, Democratization of science
2.5 Platform Architecture: The Perfect Sphere (Unified Brain Stem)
The Neurodiagnoses platform implements a "Perfect Sphere" architecture to ensure scientific integrity and prevent logic divergence. All analytical tools route through a single master endpoint:
Unified Brain Stem Endpoint:
POST /api/v2/analyze URL: https://the-alchemist-705695926498.us-central1.run.app/api/v2/analyze
This endpoint is context-aware, handling both individual patient analysis and cohort-level validation through the same underlying scientific logic:
Context 1: Individual (Clinical)
Returns individual diagnosis, confidence, causal validation score, spatial validation score.
Context 2: Cohort (Research)
Returns cohort-level statistics (t-tests, correlations), validation scores (62.5% causal, 37.5% spatial).
Critical Design: Both contexts use IDENTICAL validation logic, statistical thresholds, and biomarker mappings. This guarantees that individual and cohort analyses are scientifically congruent.
2.7 Usability & Production Readiness
Beyond scientific validation, we implemented comprehensive usability features to enable real-world clinical and research deployment:
Guided Onboarding (Epic 30)
- • 3-question wizard routes users to appropriate tools
- • Clinicians → Case Analyzer (individual patients)
- • Researchers → Data Lab (cohort validation)
- • Automated tool recommendations based on data availability
Template-Based Data Ingestion (Epic 28)
- • CSV templates with exact column headers required by backend
- • Client-side validation prevents format errors before submission
- • Templates:
proteomics_template.csv(9 columns),imaging_template.csv(9 brain regions) - • Download links embedded in upload workflows
Quality Assurance
- • End-to-end testing with real patient data (
test_case_analyzer_e2e.py) - • Automated congruence testing (
test_scientific_congruence.py) - • Zero-error CSV upload rate achieved through template + validation system
Result: Platform is production-ready for both clinical deployment and research use, not merely a proof-of-concept.
3. Results (Summary)
Causal Validation
62.5%
Genetics (Axis 1) → Molecular (Axis 2)
- • 5/8 biomarkers significant (p<0.05)
- • Tau pathology: d=1.00, p<0.001
- • Amyloid plaques: d=1.35, p<0.001
- • Phospho-tau: d=0.58, p=0.018
- • API vs Script: 0.0% deviation
Spatial Validation
37.5%
Molecular (Axis 2) → Structural (Axis 3)
- • 12/32 correlations significant (p<0.05)
- • Tau × Inferior Temporal: r=-0.602, p<0.001
- • Tau × Middle Temporal: r=-0.497, p=0.003
- • Region-specific molecular-volume associations
- • API vs Script: 0.0% deviation
Perfect Sphere Congruence Test
Automated CI/CD test verifies that individual and cohort analysis paths produce congruent results:
- • Same patient data sent to both endpoints
- • Individual path: Binary validation (validated/not validated)
- • Cohort path: Statistical validation (62.5% / 37.5%)
- • Result: 0% logic divergence detected
📦 Complete Reproducibility Package v2.0
The full manuscript (20+ pages), supplementary materials, and code are available:
- ✓ Complete manuscript text (MD + HTML)
- ✓ Traditional scripts (02_analysis_causal_validation.py, 03_analysis_spatial_validation.py)
- ✓ Perfect Sphere guardrails (test_scientific_congruence.py, ARCHITECTURE_PERFECT_SPHERE.md)
- ✓ CSV templates (proteomics_template.csv, imaging_template.csv)
- ✓ User guides (case-analyzer-guide.md, data-lab-guide.md, studio-guide.md)
- ✓ Statistical results (table2_causal_validation_statistics.csv, table3_spatial_validation_correlations.csv)
- ✓ Figures (figure2_causal_validation.png, figure3_spatial_validation.png)
How to Cite
Menéndez-González, M., Neurodiagnoses AI Development Team (2025). Neurodiagnoses Web Platform Achieves Complete Equivalence with Traditional Biostatistical Methods for Multi-Omics Causal Validation in Alzheimer's Disease. Manuscript v2.0. Available at: https://neurodiagnoses.com/science/manuscript-v2