MOSAIC Platform Compiles Chemistry Protocols to Speed Up Drug Design

Speeding up drug discovery in the age of artificial intelligence may not require abandoning traditional scientific thinking. In fact, it may rely on something surprisingly familiar: well-organized chemistry recipes. That is the core idea behind MOSAIC, a new AI-powered platform developed by researchers at Yale University in collaboration with scientists from the U.S. research arm of Boehringer Ingelheim Pharmaceuticals. The platform is designed to help chemists navigate the overwhelming volume of existing chemical knowledge and turn it into practical, reproducible laboratory procedures.

At its heart, MOSAIC addresses one of the most persistent challenges in chemistry and drug development: the difficulty of designing and executing new synthetic molecules efficiently. While chemistry has accumulated millions of reaction protocols over decades, accessing and applying that knowledge in real laboratory settings remains a major bottleneck. Researchers often spend enormous amounts of time searching literature, adapting protocols, and troubleshooting reactions. MOSAIC aims to remove much of that friction.

What MOSAIC Actually Does

MOSAIC is an AI framework that generates detailed experimental procedures for chemical synthesis, including for molecules that do not yet exist. Instead of merely predicting whether a reaction might work, it provides step-by-step guidance that chemists can realistically follow in the lab. This includes suggested reaction conditions, sequences, and procedural details derived from vast chemical datasets.

The project was led by Victor Batista, the John Gamble Kirkwood Professor of Chemistry at Yale, who also serves as a member of the Energy Sciences Institute and the Yale Quantum Institute, and as director of the Center for Quantum Dynamics on Modular Quantum Devices. Batista and his colleagues describe MOSAIC as a way to transform decades of scattered chemical knowledge into actionable laboratory intelligence.

A Different Approach to AI in Chemistry

Most existing AI tools in chemistry rely on a single, large language model to assist users. MOSAIC takes a fundamentally different route. It is powered by 2,498 individual AI “experts,” each representing deep knowledge in a specific niche of chemistry. These experts collectively cover a wide range of chemical reactions, methodologies, and experimental contexts.

The idea is similar to consulting multiple specialists rather than relying on one generalist. Just as a chef might seek advice from different culinary experts for sauces, spices, and cooking temperatures, MOSAIC draws on thousands of specialized models to assemble a complete synthesis plan. This collective intelligence approach allows the system to handle diverse chemical spaces more effectively.

Why This Matters for Chemists

Chemists typically work by following recipes, refining them through experience and experimentation. MOSAIC embraces this reality rather than trying to replace it. By making protocols easier to find, compare, and adapt, the platform reduces the time spent on trial and error and increases the likelihood that experiments succeed on the first attempt.

One of the key advantages of MOSAIC is that it includes measurable uncertainty estimates. These estimates indicate how closely a user’s request aligns with the experience of a given AI expert. In practical terms, this helps chemists prioritize experiments that are more likely to work, saving time, materials, and effort.

Demonstrated Real-World Results

The research team did not limit MOSAIC to theoretical benchmarks. They used the platform to successfully synthesize more than 35 previously unreported compounds, proving that the system can deliver real-world results. These compounds span a wide range of applications, including pharmaceuticals, catalysts, advanced materials, agrochemicals, and cosmetics.

According to the researchers, MOSAIC outperformed commercial large language models on comparable chemistry tasks. The key reason is its ability to pull expertise from thousands of narrowly focused domains rather than relying on generalized knowledge.

Who Built MOSAIC

In addition to Victor Batista, the project involved Timothy Newhouse, a professor of chemistry at Yale and co-corresponding author of the study. Newhouse emphasized that MOSAIC makes synthetic chemistry more accessible by simplifying access to protocols, much like modern AI tools have simplified searching for recipes in everyday life.

The first authors of the study are Haote Li, a former Ph.D. student in Batista’s lab who completed his doctorate at Yale in 2025, and Sumon Sarkar, a postdoctoral fellow in Newhouse’s lab. Their work played a central role in developing and validating the platform.

Open-Source and Future-Proof by Design

Another notable feature of MOSAIC is that it is fully open-source. This means researchers around the world can examine, adapt, and build upon the system without restrictions. The framework is also designed to be compatible with future AI models, ensuring that it can evolve as artificial intelligence technology advances.

The team behind MOSAIC sees it as a step toward moving AI beyond prediction and into direct support of experimental science. Instead of simply suggesting what might happen, MOSAIC helps scientists decide what to actually do next in the lab.

How MOSAIC Fits into the Bigger Picture of AI in Chemistry

Chemistry has undergone several major transitions over the years, moving from handwritten lab notebooks to printed journals, then to searchable digital databases. MOSAIC represents the next stage in that evolution: AI-guided navigation of chemical knowledge.

Rather than replacing chemists, MOSAIC acts as a smart guide, helping them make sense of enormous datasets and translate them into precise, reproducible procedures. In many ways, it functions like a digital cookbook combined with a navigation system, showing not just where to go, but how likely a given route is to succeed.

Why Faster Drug Discovery Matters

Drug discovery is notoriously slow and expensive. Developing a new drug can take over a decade and cost billions of dollars, with chemical synthesis often forming a critical bottleneck. Tools like MOSAIC have the potential to shorten development timelines, reduce costs, and enable researchers to explore chemical spaces that were previously impractical.

By accelerating early-stage synthesis, MOSAIC could help pharmaceutical researchers test more ideas faster, increasing the chances of finding effective treatments. Its applications extend beyond medicine into materials science, agriculture, and consumer products, highlighting its broad potential impact.

Looking Ahead

The developers of MOSAIC emphasize that this is not the final word in AI-assisted chemistry, but rather a foundation. As more reaction data becomes available and AI models continue to improve, systems like MOSAIC could become indispensable tools in laboratories worldwide.

The key takeaway is that innovation does not always mean abandoning old ideas. Sometimes, the smartest solutions come from reimagining familiar concepts—in this case, the humble chemistry recipe—through the lens of modern AI.

Research Paper:
https://www.nature.com/articles/s41586-026-10131-4