InterceptIQ™ Platform
A molecular grammar for early disease detection.
InterceptIQ unifies ultra-deep cell-free DNA sequencing, fragmentomic and methylation analysis, and machine-learned tissue-of-origin inference — engineered to detect active disease biology before clinical signal.
InterceptIQ™ Platform
Cell-free DNA · Molecular network · AI intelligence — read from a single tube of blood.
How InterceptIQ works
From a single tube of blood to actionable molecular intelligence.
Six tightly coupled stages convert routine plasma into a CLIA-grade readout of active disease biology. Scroll to walk the pipeline.
Blood Sample
A single standard 10 mL EDTA tube. No imaging, no biopsy, no specialized collection.
- Volume
- 10 mL
- Visit
- Outpatient
Liquid Biopsy
Centrifugation separates plasma from cellular components within minutes of draw, preserving fragile cell-free signal.
- Yield
- 4–6 mL plasma
- Time-to-process
- < 4 hrs
Cell-Free DNA
Sub-nanogram cfDNA fragments — released by dying cells across every tissue — are captured and prepared for deep sequencing.
- Input
- < 1 ng cfDNA
- Fragment size
- ~166 bp
Epigenetic Analysis
30–120× whole-genome bisulfite sequencing resolves methylation, fragmentomic, and end-motif signatures unique to each tissue and disease state.
- Depth
- 30–120×
- Sites profiled
- 28M CpG
Molecular Intelligence
Multi-task models trained on hundreds of thousands of prospectively collected samples infer tissue-of-origin and active disease biology from 2.4M features per sample.
- Features
- 2.4M / sample
- Sensitivity
- 94% at stage I
Actionable Clinical Insights
Physician-facing reports surface tissue-of-origin, disease state, and confidence — backed by analytical and clinical validation, in language clinicians act on.
- Turnaround
- 10 days
- Report format
- CLIA-grade
The interception moment
Detected at the molecular signal — years before symptoms.
Disease begins as cellular injury, becomes molecular change, then biomarker shift, then symptom. InterceptIQ™ reads the molecular signal at the moment of cellular injury — long before any clinical test would register a result.
Disease interception window
Where InterceptIQ™ sees disease that conventional testing cannot.
The natural history of Type 1 Diabetes plotted along its molecular and clinical timeline. cfDNA signal rises the moment β-cells begin to die — years before HbA1c moves.
Homeostasis
Autoimmune attack begins
cfDNA signal rises
Autoantibodies appear
Hyperglycemia, polyuria, weight loss
HbA1c, fasting glucose, OGTT
InterceptIQ™
cfDNA β-cell methylation
Years 1–3 before symptoms
- 01Healthy State
Pancreatic islet β-cells maintain insulin secretion. No autoimmune activity, no measurable tissue injury.
- 02Cellular Injury
T-cell infiltration triggers β-cell apoptosis. Dying cells release fragments of methylated DNA into circulation.
- 03Molecular Changes
Unmethylated INS gene cfDNA fragments and β-cell-specific methylation marks accumulate in plasma — years before glucose dysregulation.
- 04Biomarker Changes
GAD65, IA-2, and ZnT8 autoantibodies become detectable. β-cell mass loss accelerates past 50%.
- 05Symptoms
Clinical symptoms emerge once functional β-cell reserve is largely exhausted. Patient presents to primary care.
- 06Clinical Diagnosis
Standard-of-care diagnosis confirms T1D. By this point, >80% of β-cell mass is irreversibly lost.
Illustrative T1D natural history. Lead-time estimates supported by Akirav et al., Herold et al., and ongoing prospective InterceptIQ™ cohorts. Indicative — not for diagnostic use.
Sample → Signal → Intelligence
From a single tube of blood to clinical decision — in six instrumented stages.
The InterceptIQ pipeline is a continuous instrument loop. Every stage is quality-controlled, traceable, and engineered to preserve fragmentomic detail from picogram-level input through clinical reporting.
The signal
Every dying cell leaves a fingerprint in the bloodstream.
Capture
Proprietary low-input library chemistry preserves fragmentomic detail from <1 ng of cfDNA.
Sequence
Deep whole-genome bisulfite sequencing at 30×–120×, scaled across hundreds of thousands of samples.
Interpret
Federated AI infers tissue-of-origin and disease state from 2.4M features per sample.
InterceptIQ™ platform architecture
A vertically integrated diagnostics stack.
Five tightly coupled layers — wet lab to clinical report — each instrumented, quality-controlled, and engineered for regulatory and payer review.
Multi-site methylation architecture
Three independent epigenetic signals.
One calibrated intelligence score.
Rather than relying on a single biomarker, InterceptIQ™ interrogates multiple tissue-specific methylation loci across the insulin gene. The joint signal raises specificity, lowers false-positive rate, and produces a clinically interpretable disease intelligence output.
Open chromatin in pancreatic β-cells; hypermethylated in non-β tissue.
β-cell specific demethylation; conserved across human islet donors.
Independent confirmatory locus; lowers false-positive rate vs. single-site assays.
Calibrated weighted combination — auditable, interpretable, and reproducible across cohorts.
AI biomarker discovery
Models that learn biology, not noise.
Our models are trained on prospectively collected, IRB-approved cohorts and audited for confounding, leakage, and demographic generalization. Every classifier ships with an interpretability layer that maps predictions back to biological features — critical for regulatory review and clinician trust.
See peer-reviewed methods →Live console
What clinicians actually see — a calibrated, auditable signal.
Sample KH-PLT-0421 · synthetic readout for illustration
Tissue-of-origin
Methylation atlas
47 cell types
Fragmentomic readout
cfDNA · bp
Methylation grid · chr11