For research
Ask bigger questions of thousands of genomes.
Pangenome AI applies transformer architectures to pangenomic data, so researchers can ask questions of thousands of genomes at once instead of one reference at a time.
A single reference genome describes one individual. Biology happens across the whole population.
We help researchers decode complexity and drive breakthroughs across genetics, healthcare and bioscience.
By applying transformer architectures to pangenomic data, our Large Pangenome Models reveal genetic compatibility rules, evolutionary constraints and predictive patterns that would otherwise stay hidden across millions of genomes.
How we support researchers
5 ways in
Pangenome-scale analysis
Move beyond single reference genomes to capture the complete genetic repertoire of a species — core, accessory and everything in between.
Compatibility prediction
Identify which gene combinations work together, and which do not, before committing time and budget to wet-lab experiments.
Antimicrobial resistance
Map the genetic landscape of resistance across strains and anticipate the combinations most likely to emerge next.
Synthetic biology
Build compatibility roadmaps for designing engineered organisms that hold together outside the incubator.
Custom collaborations
Bespoke models trained on your data and pointed at your research questions, run alongside your existing programme.
What a pangenome model sees
live · alignment across strains
01 · Variation
Thousands of genomes per species, aligned so that presence, absence and rearrangement are all first-class signal.
02 · Constraint
The model learns which combinations of genes co-occur, and which the historical record has never tolerated.
03 · Prediction
Those constraints become testable predictions — compatibility, resistance, and the limits of engineering.
Bespoke models
Train on your collection, your species, your question. We build models around existing programmes rather than asking you to restructure around ours.
Discuss a collaborationData partnerships
If you hold genomic collections that deserve more than a reference-based analysis, we would like to hear about them.
Talk about dataGet involved