BEACON-1

State-of-the-art ADMET models

Beacon-1 learns from large datasets how best to represent chemical structures in order to generate highly accurate ADMET predictions.

Book a Demo
Overview of Compound 04 with chemical structure, molecular property stats including MW 389.09, HBD 1, HBA 4, TPSA 72.47, RBN 6, Fsp3 0.59, and ADME parameters such as LogD 0.39, Kinetic Solubility 160, HLM 1.8, MDCK-MDR1P app 6.6, MDCK-MDR1 ER 2.9, and hPPB 87%.

Consecutive 1st place finishes in the world's largest ADMET benchmarks

Graphic announcing OpenADMET's blind competition win in 2026 against ExpansionRx, showing rankings with inductivebio first, Merck with Nvidia second, EMD Serono third, and 370+ other competitors; includes statistical metrics and two chemical structures.

1st place in OpenADMET’s 2026 blind ADMET competition with ExpansionRx

Inductive’s Beacon-1 models won the OpenADMET and ExpansionRx blind challenge - a real-world ADMET competition featuring more than 370 submissions to predict up to 9 endpoints across over 2,250 compounds.

Explore the winning approach
2025 Winner OpenADMET's blind competition with ASAP, ranking first inductivebio, second Galápagos, and 37 other competitors, with statistics on Mean Absolute Error 0.224, Mean Squared Error 0.104, Pearson R 0.802, and R² 0.623, along with two molecular structures.

1st place in OpenADMET’s 2025 blind ADMET competition with ASAP

Inductive’s Beacon-1 model won the Polaris Ligand ADMET competition – 
A blind, open competition of 39 competitors from leading AI drug discovery companies and academic groups focused on predicting ADMET for a coronavirus main protease program.

See how we did it

Models available:

Microsomal Stability
Hepatocyte Stability
Cell Membrane Permeability
Efflux
Solubility
Experimental LogD
Plasma Protein Binding
pKa
hERG Inhibition
In Vivo PK
BBB
Off-target Toxicity
DDIs

Every model is fine-tuned & regularly updated to ensure optimal performance in your chemical space

Cluster of light blue particles or cells spreading outward on a black background, forming a porous, organic pattern.
Example table

Heterogenous data are standardized for training, and Beacon-1 learns to adjust for variance across each assay

CalloutCalloutMobile