Technical Overview

How Diadia's AI Analyzes
Your Biomarkers — And Why
It's Different

A structured, multi-stage clinical investigation that catches what standard interpretation misses.

The Multi-Stage Pipeline
Seven distinct stages transform raw lab data into a personalized, evidence-backed protocol. Each stage builds on the last — not a single-pass guess.
01
Input

Dual-Range Biomarker Evaluation

Every marker is scored against both the conventional lab range and a tighter optimal range derived from healthy-population research. Subclinical dysfunction — the grey zone — gets flagged, not ignored.

02
Analysis

Independent Biomarker & Symptom Analysis

Two parallel tracks: biomarker patterns identify potential root causes through multi-marker fingerprinting, while symptoms and health history independently surface additional candidates. No confirmation bias.

03
Synthesis

Cross-Stream Convergence

Causes supported by both biomarkers and symptoms are strengthened. Single-source causes are critically evaluated — kept if strong, deprioritized if weak. Like a diagnostic team conferring.

04
Genetics

Targeted Genetic Querying

Causes first, genetics second. Only the 50–100 clinically validated SNPs relevant to identified causes are examined — not all 700K+ variants. Signal, not noise.

05
Hierarchy

Root-Cause Hierarchy & Ranking

Each cause is placed in a physiological hierarchy — upstream neuro-endocrine-immune, metabolic systems, or foundational nutrient level — then ranked by evidence weight and patient priorities.

06
Verification

Mechanistic Research Verification

Every recommendation is verified via mechanistic chain reasoning against published literature. Claims that fail are excluded. Citations trail back to real studies — no hallucinations.

07
Output

Tiered, Complaint-Organized Protocol

Foundation-tier items target your primary concerns directly. Support-tier addresses cofactors. Each item includes specific dosing, timing, rationale, and how to obtain it.

Optimal vs. Reference Ranges
The reference range tells you if you're sick. The optimal range tells you if you're healthy. Diadia evaluates both — catching the subclinical grey zone.
TSH (Thyroid Stimulating Hormone)
Patient value: 4.2 mIU/L
0.0 Optimal: 1.0–2.5 5.0+
IN LAB RANGE · OUTSIDE OPTIMAL → Early thyroid dysfunction flagged
Ferritin (Iron Storage)
Patient value: 25 ng/mL
0 Optimal: 50–100 200+
IN LAB RANGE · WELL BELOW OPTIMAL → Iron depletion explains fatigue & hair loss
Vitamin D (25-OH)
Patient value: 62 ng/mL
0 Optimal: 50–80 100+
IN LAB RANGE · IN OPTIMAL RANGE → No action needed
Lab Reference Range
Optimal Range
Optimal
Suboptimal
Functional Imbalance
Independent Analysis, Then Synthesis
Two parallel streams form independent impressions — like a radiologist and clinician — before converging. This prevents confirmation bias and catches hidden patterns.
Biomarker Analysis
Insulin Resistance
↑ fasting glucose · ↑ HOMA-IR · ↑ triglycerides
Thyroid Dysfunction
TSH 4.2 · low-normal fT3 · ↑ rT3
Iron Depletion
ferritin 25 · ↓ iron saturation
HPA Axis Dysregulation
cortisol pattern flat · DHEA-S low
Symptom Analysis
Insulin Resistance
afternoon crashes · sugar cravings · weight gain
Iron Depletion
fatigue · hair loss · brittle nails
Gut Dysbiosis
bloating · food sensitivities
Sleep Disruption
waking 3am · unrefreshing sleep
Synthesis Result
Strengthened: Insulin Resistance & Iron Depletion — supported by both streams.
Kept (biomarker-only): Thyroid Dysfunction — strong lab evidence, no symptoms yet.
Deprioritized: Gut Dysbiosis — symptom-only, no lab confirmation.
Targeted Genetics — Not a Haystack
Consumer genetic tests yield 700K+ data points. Scanning all of them produces noise. Diadia queries only the validated variants relevant to your specific identified causes.

Brute-Force Approach 700K+ VARIANTS

Every variant scanned. Thousands of "risk" flags. No clinical context.

Each square = ~500 variants. Red = flagged "risk" with no clinical context.

Diadia's Approach 50–100 TARGETED SNPs

Causes identified first → only relevant, validated variants queried.

MTHFR Methylation impairment → thyroid dysfunction
DIO2 T4→T3 conversion → explains low fT3
HFE Iron metabolism → explains low ferritin
TCF7L2 Insulin signaling → insulin resistance risk
COMT Catechol metabolism → HPA axis regulation
Physiological Hierarchy
Causes are organized by where they sit in the body's causal chain — upstream drivers that create downstream effects. Treatment targets the source.
Tier 1 · Upstream
Neuro-Endocrine-Immune
Widest downstream impact — treat first
HPA Axis Dysregulation
Thyroid Dysfunction
Chronic Inflammation
Immune Dysregulation
drives
Tier 2 · Metabolic
Organ & System Level
Bridges upstream drivers to downstream effects
Insulin Resistance
Gut Dysbiosis
Hepatic Dysfunction
Mitochondrial Dysfunction
drives
Tier 3 · Foundational
Nutrient & Cellular Level
Often consequences of upstream dysfunction
Iron Depletion
Vitamin D Insufficiency
Magnesium Deficiency
Oxidative Stress
Mechanistic Chain Reasoning
No single trial may test an exact combination. Diadia decomposes claims into biological links, verifies each independently, then validates the chain — exactly how an experienced clinician reasons.
Example: Methylfolate for MTHFR + Impaired Methylation
MTHFR variants impair folate metabolism
✓ Strong evidence
Impaired folate reduces methylation capacity
✓ Biochemical evidence
Methylfolate bypasses MTHFR to restore methylation
✓ Mechanistic + clinical
Claim validated via chain reasoning. Each link independently supported by published research. No single trial needed for the full chain — and no hallucinated citation.
⟶
Direct Mechanism
A directly causes B, supported by a published study
⟶⟶
Chain Reasoning
A→B and B→C each evidenced, so A→C logically established
⇉
Pathway Convergence
Condition X disrupts pathway P, intervention Y modulates P → Y addresses X
Diadia vs. General AI
A structured clinical investigation vs. a single-pass chatbot response.
Diadia's Analysis
General AI / Single-Pass
Range Interpretation
✓ Optimal + lab ranges (catches subclinical)
✗ Lab reference range only
Analysis Approach
✓ Independent streams → synthesis
✗ Single pass, all data at once
Genetic Integration
✓ Cause-targeted, 50–100 SNPs
✗ All 700K variants (noise) or none
Cause Identification
✓ Root-cause hierarchy
✗ Surface-level pattern matching
Recommendations
✓ Domain-specific specialist agents
✗ Generic suggestions, one model
Evidence Verification
✓ Mechanistic chain reasoning + citations
✗ No verification (hallucinated citations)
Output Format
✓ Tiered protocol with dosing
✗ Unstructured text