DNAI
Research use only. Not approved for clinical decision-making or diagnostic procedures.
For Clinicians

Simulate Treatment Scenarios. Compare Molecular Responses.

DNAI creates patient-specific digital twins from multi-omics data to simulate therapy outcomes — helping researchers explore treatment scenarios with quantified uncertainty and generate testable hypotheses.

Request Demo
Primary Use Case

The Virtual Tumor Board

A dedifferentiated liposarcoma patient faces standard-of-care doxorubicin. The researcher needs to explore: would a CDK4/6 inhibitor combination show higher simulated sensitivity?

Step 1

Molecular Profiling

Input
Patient biopsy: RNA-seq + DNA mutations + CNV + Methylation + Histopathology (WSI)
Matched WSI required for BRCA and other WSI-dependent types (C=0.37 omics-only); RNA-seq minimum via probabilistic fusion
VAE v5.10 encodes multi-omics into 328-dimensional structured latent space (C-index: 0.704, internal val, survival ranking)
Driver Module identifies CDK4 amplification (driver probability: 0.85) and MDM2 co-amplification via pathway analysis
Signal Detected
CDK4/MDM2 co-amplification detected. Cell cycle dysregulation dominant. Simulation suggests CDK4/6 inhibitor + MDM2 inhibitor combination may show higher sensitivity in this molecular context.
ILLUSTRATIVE — based on simulated patient profile
Step 2

The Simulation

Scenario A: Doxorubicin (SoC)
6-month projectionTumor continues growing
Simulated responseProgressive Disease
Simulated: LOW SENSITIVITY
Scenario B: Abemaciclib + Idasanutlin
6-month projectionTumor reduction
Simulated responsePartial Response
Simulated: HIGH SENSITIVITY
Structured Abstention Layer
If patient data is out-of-distribution or information is insufficient, the system flags: "Uncertainty too high — results not reliable for this patient." and explains what data is missing.
Step 3

The Research Insight

1
Simulation shows low sensitivity signal for chemotherapy in this molecular context
2
Pathway analysis traces response difference to CDK4/MDM2 amplification — traceable, not black-box
3
Researchers can use this insight to inform treatment planning alongside standard clinical evidence
Research Value
Explore treatment hypotheses with quantified uncertainty
Every simulation traces to specific pathways and genes
Model uncertainty intervals and abstention flags included

How DNAI supports clinical research

Simulation-informed insights for treatment planning and research

Rapid simulation

Get simulated treatment outcomes in seconds. Upload multi-omics data and receive immediate response estimates with model uncertainty intervals.

Patient-specific

Each simulation is derived from the individual patient's molecular profile — not population averages.

Uncertainty quantified

Model uncertainty intervals and abstention flags accompany all simulations. The system tells you when not to trust a result.

Research applications

Treatment scenario comparison

Rank treatment options by simulated sensitivity for each patient. Compare targeted therapies, immunotherapy, and combination regimens with model uncertainty intervals.

Survival risk modeling

Simulate survival trajectories to explore prognosis hypotheses. Per-horizon calibration at 1yr, 2yr, 3yr, and 5yr (internal validation).

Resistance scenario simulation

Simulate tumor evolution to explore potential resistance emergence patterns and generate hypotheses about adaptive treatment strategies.

See the Clinical Platform

Explore the research tools for patient-specific simulation

Patient Dashboard

Patient Dashboard

Overview of all patients

Platform Tour

Platform Tour

Full pipeline walkthrough

Model Architecture

Model Architecture

Pipeline explorer

Validation

Validation

Calibration metrics

Explore full platform tour

Known Limitations

15 of 33 cancer types have C-index below 0.60 — structured abstention recommended for these types
BRCA and other WSI-dependent cancer types show near-random performance without histopathology (C=0.37 omics-only)
Drug response and resistance are simulated from molecular profiles — not clinically validated against treatment outcomes
External validation degrades with platform shift (microarray vs. RNA-seq) — RNA-seq data recommended

Intended use

DNAI is intended solely as a research tool for hypothesis generation and exploratory molecular analysis. It is not approved, cleared, or validated as a clinical decision support system, diagnostic device, or companion diagnostic. All simulations should be interpreted alongside standard clinical evidence by qualified researchers.

Schedule a research demo

See how DNAI simulation tools can support your oncology research.

Request Demo