Platform · Prediction & modeling
One model, 150,000+ endpoints, run in both directions.
Pleiome is the generative successor to ToxTransformer, in development. It predicts 150,000+ ADMET, toxicological, and biochemical endpoints in a single pass — 22.5x ToxTransformer's coverage — trained on the entire Yard knowledge graph rather than a subset. Run it forward to score a molecule, or backward to design one: give it a target safety profile and it generates novel, synthesizable structures.
Updated July 18, 2026
- 150,000+
- endpoints predicted at once
- 22.5x
- the endpoint coverage of ToxTransformer
- 0.942
- mean AUROC on the ToxBench core tasks
Capabilities
What Pleiome does
Predicts broadly
One transformer covering 150,000+ ADMET, toxicological, and biochemical endpoints in a single pass, each with a confidence you can weigh.
Design, don't just screen
Run in reverse: give it a target profile and it proposes novel, synthesizable structures, optimizing multiple objectives at once.
Trained on the whole graph
Learns from the entire Yard knowledge graph, giving it far more signal than the public data most models see.
How it works
From question to result
- 1
Predict
Score a molecule across 150,000+ endpoints in a single pass, each with a confidence you can weigh.
- 2
Compare
Read predictions beside measured data from Kiln and automated OECD read-across from ToxJobs.
- 3
Design
Invert the model to generate candidate structures that meet a target ADMET and toxicity profile.
Related
Works closely with
FAQ
Pleiome questions, answered
Frequently asked questions
What is Pleiome?
Pleiome is a generative transformer trained on the Yard knowledge graph that predicts 150,000+ endpoints at once and can run inverse design to hit a target safety profile. It is in development.
How much more does Pleiome cover than ToxTransformer?
Pleiome predicts 150,000+ endpoints at once — 22.5x ToxTransformer's coverage — while holding 0.942 mean AUROC across the ToxBench core tasks.
What is inverse molecular design?
Instead of scoring a molecule you already have, you give Pleiome a target ADMET and toxicity profile and it generates novel, synthetically accessible structures that meet it, optimizing several objectives at once.
See it on your own chemistry.
Try the free version now, or book a walkthrough of the full platform on your substances.