NDx

PRECISION NEUROLOGY

AI-enabled multi-omic blood tests for neurodegenerative diseases

We turn a routine blood draw into a real-time readout of brain health.
So drug developers enroll the right patients and learn sooner whether a therapy works.

Ze Yang CATALYST · SEPT 2026

The problem

The measurement gap in neurodegenerative diseases

— hampers drug development, clinical monitoring, and ultimately patient care.

The scope

57M

living with dementia · ~10M new cases/yearWHO

1,023

ongoing ND drug trials · 678 in Phase 2/3CT.gov

$1.3T

annual global cost of dementiaWHO

How we measure today

Amyloid PET

~$5,000 / scan

Too expensive and scanner-boundUSC

A lumbar puncture performed through a sterile drape

CSF / lumbar puncture

75% → 64%

Invasive — drops enrollment willingnessPMC

A hand holding several blood collection tubes

Plasma pTau217

1 analyte

Amyloid only, AD only — doesn't generalizeFDA

Who pays for it

Drug developers

>70% screen-fail, then 12–18 mo to a clinical-scale readoutAlz&Dem

Clinicians

3 in 4 dementia cases go undiagnosed — and no objective way to monitorADI

Patients

Months of waiting before a symptom scale can register whether anything has changed

One root cause across every disease: no scalable, longitudinal readout of brain biology.

Images via Wikimedia Commons — PET: U.S. Navy, public domain · LP: Dragondefuego1976, CC BY-SA 4.0 · Tubes: Tannim101, CC BY 3.0

The solution

AI-powered blood diagnostics for precision neurology

Decoding multi-omic brain signals to stratify patients, track progression, and inform treatment development.

Start

Plasma sample

A simple blood draw — universally accessible, repeatable over time

Two complementary signals

cfDNA methylation

Direct molecular readout of brain-cell death — which tissue is dying, right now

Autoantibody profile

Immune response to that damage — biologically amplified, patient-specific

AI integration · two-fold

Agentic workflow

Runs the analysis end to end — an intelligent, efficient and generalizable flow from raw assay data to the finished report

Multimodal AI model

Consumes the multi-omic data together and turns it into diagnostic scores — more than either assay carries alone

Unified platform

Patient stratification

Who should enroll, and how should patients be grouped biologically

Progression monitoring

How disease biology is changing during treatment, sample to sample

First-principle biological foundation with platform potential

Pilot study data

cfDNA methylation measures brain degeneration directly at the molecular level

4 CBS participants + 3 healthy controls · plasma cfDNA methylation (5mC) profiling

Significant brain-derived signal

Broad biomarker discovery

Biomarker with biological resolution

Brain-derived cfDNA fraction, CBS versus healthy controls

CpG-level differential methylation volcano plot, CBS versus healthy controls

CBS cfDNA 5mC brain circuit map across 90 brain-fraction DMR genes

CBS: corticobasal syndrome.

Pilot study data

Autoantibody profiles measure immune heterogeneity and brain degeneration indirectly

Shown: 13 AD participants + 9 healthy controls · IgG/IgM autoantibody profiling

Discover candidate biomarkers

Reveal patient-level variation

Compare disease-associated profiles

Clustered heatmap of 137 confirmed autoantibody hits, AD versus healthy controls

Personalized autoantibody profiles across AD patients

PCA of AD versus healthy-control autoantibody profiles

Overall study: 60 samples across AD, PD and healthy controls.

Why both

Two measurements combined give the full picture

Complementary biology, complementary timescales.

cfDNA methylation

Brain degeneration

Direct readout of dying brain cells

Extreme analytical sensitivity

Real-time monitoring

Limited historical context

Autoantibody profile

Immune response

Biological amplification of weak signal

Personalized disease fingerprint

Persistent immune signature

Limited timing precision

Combined approach

One integrated report per patient, per timepoint

NDx Multi-omic Patient Report PT‑0143 · VISIT 3 · ILLUSTRATIVE

Brain readout · cfDNA

Brain fraction 2.6%

Cortex ↑ 1.9×

Striatum / BG ↑ 1.3×

Cerebellum n.s.

Longitudinal monitoring

AI diagnostic scores

78 /100 Alzheimer's disease High

34 /100 immune activity Moderate

AD subtype likelihood

Limbic-pred. 0.61

Typical 0.25

Hipp.-sparing 0.14

Model v1.2 · locked, versioned, traceable

Clinical context

68 F · MCI, first referral 2025-08

History: hypertension; family history of AD

Cognitive: MoCA 23 → 21 over 12 months

Trial fit: meets biological entry criteria

Neither assay alone answers a pharma question. The paired readout is the product.

Pharma applications

One blood-based platform.
Two critical trial decisions.

For neurodegeneration drug developers.

1

Patient selection
& stratification

Who should enroll, and how should patients be grouped?

Intended benefits

More targeted screening

Biologically informed patient groups

Lower screening burden and cost

Decision it serves — who enters the trial

2

Longitudinal
monitoring

How is disease biology changing during treatment?

Intended benefits

Repeatable molecular measurements

Potential earlier treatment-response signals

Better-informed go/no-go decisions

Decision it serves — is the therapy working

Proposed applications; clinical utility and longitudinal performance to be validated.

Competitive advantages

NDx's competitive edge in precision neurology

	NDx	Other multi-omic platforms	PET / CSF	Single-protein blood marker
Proprietary multi-disease data	✓	✓	Limited	Limited
Dual biological readout epigenetics + immune	✓	✕	✕	✕
AI integration	✓	✓	✕	✕
Pan-neurodegeneration potential	✓	Varies	✕	Varies

Only NDx pairs two orthogonal biological readouts from a single blood draw and integrates them with a model trained on its own multi-disease data.

Our current state

Human proof-of-concept completed —
ready for validation

~60-patient, 2-year pilot completed

Initial multi-omic human dataset generated

Core 5mC + autoantibody workflows established

Automated AI-assisted pipeline established

Stanford biobank collaborations

Stanford OTL disclosure + docket

Established assay-to-report workflow

Signal extraction is demonstrated in human plasma. Next is validation at scale.

Catalyst milestones

Catalyst milestones for a pharma pilot

Milestone

Success criteria

I

Establish paired value

AI integrates cfDNA methylation + autoantibodies. The locked model adds value over either assay alone in held-out participants.

II

Establish longitudinal relevance and reliability

Biomarker changes associate with clinical change. Replicates and batches meet predefined QC criteria.

III

Prepare a defined pharma pilot

Prioritize the first indication and use case. Deliver an assay-to-report workflow with a locked AI model, evidence package and pilot protocol.

Proposed study: 160 participants × 2 timepoints = 320 plasma samples.

40 per group — AD · PD · ALS · controls Baseline + 12 months Both assays per sample

12-month execution plan

Concrete 12-month plan to build and validate the prototype

MONTH

012 345 678 91011 12

Paired dataset and assay QC

~~M0–6~~

Confirm banked samples and clinical metadata

Profile both assays, measure repeatability

Paired value and longitudinal relevance

~~M4–9~~

Develop and validate the AI integration model

Test added value and clinical-change associations

Prepare a defined pharma pilot

~~M8–12~~

Lock the assay-to-report workflow and AI model

Finalize first use case, evidence and pilot protocol

AI Research Assistant prototype QC and candidate prioritization

Pharma discussions shape the use case and pilot requirements

Deliverable

QC-qualified paired
dataset AI validation report +
first application Locked model + pilot
evidence package

The business model

Integrated Precision Medicine
Business Flywheel

1

Pharma servicesAccelerate trial enrollment · upfront + per-sample fees

2

Proprietary dataIn-house multi-omic pipeline · data licensing + insights

3

CDx partnershipsCo-develop targeted assays · clinical + regulatory milestones

4

Clinical testingFDA-cleared rollout to neurologists · recurring reimbursement

Accelerated Growth FlywheelEvery sample sharpens the next readout.

Revenue funds data · data improves the model · a better model wins the next contract.

Team

Built to take biomarkers from
discovery to clinics

Wet lab, computational biology and ML product — with the clinical access to run the study.

CEO

Ze Yang

Multi-omics & wet lab

Scientist, Stanford University
PhD, Stanford University

CTO

Daniel Zheng

AI & Product

Siri AI Engineer, Apple
Ex-CTO, Utopia Studios

CSO

Siyu He

AI & computational biology

Postdoc, Stanford University
PhD, Columbia University

Advisor

Kathleen Poston

Clinics & biomarkers

Professor of Neurology
Stanford Medicine

Our ask

Funding and expertise for our first pharma pilot

Proposed 12-month project budget

Multi-omic profiling $400K 40%

Project personnel $360K 36%

AI scientist development + cloud $115K 11.5%

Validation, regulatory, QC & translation $125K 12.5%

Catalyst support requested

Pharma connections

Reach biomarker decision-makers and shape the first pilot with partner input

Validation and regulatory guidance

Define fit-for-purpose evidence requirements and a regulatory roadmap for the intended use

Commercialization support

Refine the business model and OTL licensing path, and prepare for follow-on funding

WHAT CATALYST ENABLES

A reproducible blood biomarker platform prototype ready for a defined pharma pilot

Helping drug developers understand how disease biology differs between patients and changes over time.

yangze@stanford.edu

Stanford OTL invention disclosure: S26-188 · patent and licensing path to clarify with OTL