Arctoris
Accelerating drug discovery through automated, robotics-driven data generation and AI-integrated laboratories.
Accelerating drug discovery through proteome-wide AI-driven polypharmacology.
Cyclica, now a core component of Recursion Pharmaceuticals' technology stack following its 2023 acquisition, represents the frontier of biophysics-informed AI. Its architecture is built around two primary engines: MatchMaker and POEM. Unlike traditional drug discovery tools that focus on a 'one-target, one-drug' philosophy, Cyclica utilizes a 'polypharmacology' approach, screening small molecules against the entire human proteome to identify both primary therapeutic targets and potential off-target effects early in the pipeline. MatchMaker utilizes deep learning to predict protein-ligand interactions by integrating structural data and biochemical knowledge, while POEM (Property Estimation Model) provides a machine-learning-based approach to predicting ADMET (Absorption, Distribution, Metabolism, Excretion, and Toxicity) profiles. By 2026, the platform has been fully integrated into Recursion’s 'OS', serving as the primary computational engine for identifying novel chemical entities (NCEs) and drug repurposing candidates. It bridges the gap between genomic data and clinical outcomes by providing a holistic view of how molecules interact with the complex biological landscape of a cell.
A deep learning engine that combines structural proteomic data with interaction networks to predict binding affinities without requiring a crystal structure.
Accelerating drug discovery through automated, robotics-driven data generation and AI-integrated laboratories.
Accelerating drug discovery through deep physics and generative AI without experimental data training.
Augmenting human intelligence to discover and develop life-changing medicines via end-to-end AI drug discovery.
Pioneering hypothesis-free drug discovery through the Interrogative Biology® platform and Causal AI.
Verified feedback from the global deployment network.
Post queries, share implementation strategies, and help other users.
A non-linear, machine-learning-based ADMET predictor that uses fingerprint-based molecular descriptors.
Simultaneous docking of a ligand against thousands of protein structures simultaneously.
Comparing ligand binding across different species proteomes or different cell states.
A comprehensive database of all searchable binding pockets across the known proteome.
Built on Kubernetes-based architecture to handle millions of molecular simulations in parallel.
Provides heatmaps showing which atomic features contributed most to a binding prediction.
Late-stage drug failure due to unforeseen toxicity.
Registry Updated:2/7/2026
Finding new uses for existing FDA-approved drugs.
Identifying which protein a phenotypic hit is actually binding to.