Chemistry & ADMET · Sagawa & Kaneko · IBM RXN
Forward and retrosynthetic prediction, plus RXNFP fingerprints for comparing reactions.
1–30s
Runs on On-demand GPU
Runs on our servers with on-demand compute. A first run needs time to load the model; active capacity can be reused and scales down when idle.Run this model
About this model
Reaction prediction treated as translation: give it reactants and it writes the product, give it a product and it writes reactants that would make it. Trained on USPTO patent reactions, so it predicts what the literature would say — useful for checking whether a generated molecule is plausibly makeable before it reaches a chemist's queue, and not a claim that any particular step will work. Retrosynthesis returns a ranked set because a molecule has many plausible disconnections and there is no single right answer.
Standardized I/O contract
Every model in the Hub speaks the same contract, which is what lets the Router and the agent call any of them without special-casing.
Inputs
Reaction or molecule (SMILES)
Reagents and solvent (SMILES)
What to compute
Compare with
Predictions
Outputs
Ranked SMILES with confidence, each checked for whether it parses.
