Rovira i Virgili University Unveils CoCoGraph, Beating Rivals on Nearly Two-Thirds of 36 Tests
Updated
Updated · Futura · Jul 10
Rovira i Virgili University Unveils CoCoGraph, Beating Rivals on Nearly Two-Thirds of 36 Tests
1 articles · Updated · Futura · Jul 10
Summary
CoCoGraph, developed by researchers at Spain’s Rovira i Virgili University, generates millions of new molecules while enforcing chemical rules so every output is chemically valid.
The AI uses a diffusion model that learns by breaking down real molecules and rebuilding them, aiming to speed drug and materials discovery in a chemical space estimated at up to 10^60 possibilities.
Across 36 physicochemical properties, CoCoGraph’s molecules looked more realistic than rival systems on nearly two-thirds of the measures, while using fewer parameters and less computing power, the team said.
In a test with 121 chemists, experts mistook AI-generated molecules for real ones about 40% of the time, underscoring how convincing the system’s designs can be.
The Nature Machine Intelligence study adds to a fast-growing wave of chemistry diffusion models that could shorten the search for new medicines and high-performance materials.
How does this AI go from generating millions of molecules to creating one specific, synthesizable drug for a targeted disease?
Does this AI truly understand chemistry, or is it just an expert mimic that could trap discovery within the limits of existing data?
When an AI invents a breakthrough molecule, who legally owns the patent: the creator, the user, or the machine itself?
CoCoGraph: AI Model Achieves 100% Chemically Valid Molecule Generation and Powers an 8.2 Million Compound Database
Overview
CoCoGraph is a new artificial intelligence model developed by researchers at Universitat Rovira i Virgili and published in Nature Machine Intelligence in May 2026. Unlike earlier AI models that often created chemically impossible molecules, CoCoGraph adapts diffusion models—originally used for generating realistic images—and applies them to the strict rules of chemistry. By directly incorporating fundamental chemical rules like atomic valence into its process, CoCoGraph ensures that every molecule it generates is 100% chemically valid. This breakthrough overcomes previous limitations and opens new possibilities for drug discovery and materials science.