A Russian startup has developed an AI model that outperforms Nobel Prize-winning technology in one of drug discovery's most critical tasks. Ligand Pro, founded by Skoltech professors and researchers, unveiled Matcha, an artificial intelligence system that screens potential drug molecules against disease-causing proteins 30 times faster than AlphaFold3, the cutting-edge AI developed by DeepMind that won the 2024 Nobel Prize in Chemistry. The breakthrough could democratize early-stage drug development by making it feasible for smaller research centers that previously lacked the computational resources .

What Makes Virtual Drug Screening So Important?

Before scientists test a potential drug in the laboratory, they use computational methods to predict whether a drug molecule will actually bind to its target protein. This process, called molecular docking, is like fitting a key into a lock. The drug molecule must have the right shape and chemical properties to slip into a cavity on the protein called a binding pocket. Virtual screening allows researchers to test millions of compounds computationally before investing time and money in physical experiments, potentially saving years of work and millions of dollars .

In 2020, DeepMind introduced AlphaFold, an AI system that solved a decades-old problem by accurately predicting 3D protein structures. This breakthrough was so significant that its developers received the 2024 Nobel Prize in Chemistry. However, even with accurate protein structures available, the next challenge remained daunting: matching millions of drug molecules to those proteins quickly enough to be practical .

How Does Matcha Outperform AlphaFold3?

The speed difference is dramatic. Matcha processes a single protein-drug complex in 13 seconds, compared to AlphaFold3's 6.5 minutes. When screening an entire database of millions of compounds, AlphaFold3 requires approximately four and a half months of continuous computation, while Matcha completes the same task in less than eight days . This 30-fold speed advantage comes without sacrificing accuracy or reliability.

What sets Matcha apart is its approach to ensuring physical realism. The algorithm works through a multi-step process that first determines a molecule's approximate position within the protein pocket, then adjusts its rotation and internal structure. The system uses a physics-aware verification method called GNINA that automatically discards configurations that violate the laws of chemistry and physics. Only physically realistic configurations are ranked by predicted binding strength to identify the optimal match .

"Drug development is a long, capital-intensive, and high-risk process. A project can be stopped at any stage, even after significant time and effort has been invested. Even so, computational methods can optimize the early stages particularly well. Our mission is to create effective AI-based tools that would establish a comprehensive computational framework in the early stages of drug development," said Marina Pak, co-founder and CEO of Ligand Pro and a Skoltech alumna.

Marina Pak, CEO of Ligand Pro

How to Leverage Matcha for Drug Discovery

• Virtual Screening at Scale: Mid-sized research centers can now screen millions of drug candidates computationally without access to supercomputers, making early-stage drug discovery more accessible to institutions with limited budgets.
• Faster Hit Identification: By reducing screening time from months to days, researchers can identify promising drug candidates more quickly and move validated compounds into laboratory testing with greater confidence.
• Cost Reduction in Early Development: Virtual screening before physical experiments eliminates the need to synthesize and test millions of compounds, reducing material costs and researcher time in the earliest, most exploratory phase of drug development.
• Integration with Existing Workflows: Ligand Pro released the code, model weights, and manuscript openly, allowing researchers to integrate Matcha directly into their current research and development pipelines without waiting for commercial availability.

The accessibility of Matcha represents a significant shift in computational drug discovery. Previous AI breakthroughs in protein structure prediction, while revolutionary, still required enormous computational resources to apply to practical drug screening. By solving the speed bottleneck, Matcha makes virtual screening viable for research groups that previously could only afford to test a small fraction of potential candidates .

"In just three years, we've gone from proposing an idea and building a team to achieving game-changing results. We continue to develop Matcha and address related tasks, including the generation of molecules, prediction and optimization of their properties. Our next step is to validate our solution experimentally in real R&D pipelines, and then proceed with industrial implementation," explained Daria Frolova, head of machine learning at Ligand Pro and a Skoltech Computational and Data Science and Engineering PhD student.

Daria Frolova, Head of Machine Learning at Ligand Pro

The team's next phase involves testing Matcha in actual pharmaceutical research and development environments. If real-world validation confirms the computational results, the technology could accelerate the timeline for bringing new drugs to market. This matters because drug development typically takes 10 to 15 years from initial discovery to regulatory approval. Any efficiency gain in the earliest stages compounds over time, potentially bringing life-saving medicines to patients years sooner .

By making the research openly available, Ligand Pro has also enabled independent verification of their claims and broader adoption across the scientific community. This transparency approach contrasts with some proprietary AI systems and could accelerate innovation by allowing other researchers to build upon or integrate Matcha into their own tools .