Our Approach

How the generator
is built

Reference-conditioned autoregressive generation plus condition injection. A clinical caller is in development. It is not for sale.

Generator

Reference-conditioned architecture

The product you can buy is labelled FASTQ from a reference-conditioned generator, not a clinical prenatal test.

Research on low fetal fraction remains open. We do not sell a 1–2% FF clinical test.

Key technical innovations include:

  • Reference-conditioned generation so injected karyotypes survive alignment
  • Controllable fetal fraction, karyotype, and fragment depth
  • Paired-end FASTQ plus metadata for caller and pipeline regression

Data Generation

Synthetic cfDNA Technology

Our synthetic data generation uses a reference-conditioned autoregressive model (132.5M parameters) trained on real cfDNA fragments. The model generates sequences conditioned on the GRCh38 reference genome so injected karyotypes can survive alignment.

The generation process provides precise control over:

  • Fetal fraction: 1-25% with continuous control
  • Karyotype: T21, T18, T13, 45,X, 47,XXX, the rare autosomal trisomies, and the autosomal microdeletions and duplications generate the expected signal today. T21 has been checked end to end on sequenced output; the others pass our internal tests only.
  • Fragment characteristics: size distribution, GC content, coverage patterns
  • Sample depth: 4M to 32M fragments on the public pack path

This technology supplies labelled FASTQ for NIPT algorithm development. A clinical caller is in development and is not for sale.

Validation

Validation framework

This is the one validation result we claim.

Level 1

Train Synthetic, Test Real

A classifier trained only on our synthetic data reached AUC 0.87 for T21 on one published dataset (Lun et al. 2014: 26 unique samples, 6 T21 and 20 euploid; p = 0.002).

AUC 0.87 for T21

Results

Synthetic Data Validation Results

0.87AUC for T21A classifier trained only on our synthetic data reached AUC 0.87 for T21 on one published dataset (Lun et al. 2014: 26 unique samples, 6 T21 and 20 euploid; p = 0.002).
View Full Validation Report Browse Conditions

Research

Publications

Research publications are forthcoming in 2026. We are preparing manuscripts detailing generator architecture and synthetic-data experiments.

Preprint expected H2 2026.

Open Science

Commitment to Open Source

We believe in advancing prenatal genomics through open collaboration. Synthetic cfDNA generation is available now for research use from the dashboard, so labs can reproduce experiments without sharing patient data.

View our GitHub

Get labelled FASTQ

Order from the dashboard or read how we tested it.

PricingHow we tested it