Biohub's AI Breakthrough: Designing Proteins Like Writing Code

Designing Cancer Drugs with Code

Exploring the world of AI in health innovation.

Why this matters

Understanding the intricacies of protein folding and binding.

When it comes to the intricate world of protein biology, the 3D structure of a protein often dictates its function. Proteins that do not fold correctly may fail to function as intended, or worse, cause disease. Scientists have long sought to understand and predict how proteins fold, a process that can take months or years experimentally. However, recent advancements in AI have dramatically accelerated this process, potentially revolutionizing fields like cancer treatment and immunology.

The journey to AI-accelerated protein design

Understanding protein folding

Imagine being able to design a cancer drug with the same precision as writing a line of code. This once-futuristic concept is now a reality, thanks to a recent breakthrough by Biohub. For decades, scientists have grappled with predicting how proteins fold, a critical step in understanding and manipulating their behavior. A protein's 3D shape determines its function, what it binds to, and whether it fights or causes disease. Getting the shape wrong can result in a drug that doesn't work, while getting it right can lead to powerful therapeutic agents.

From prediction to design

The quest to understand and predict protein folding has seen significant milestones. AlphaFold, developed by DeepMind, made waves in 2020 by solving the protein folding problem and winning the Nobel Prize in Chemistry. AlphaFold's success, however, was primarily in prediction. The more challenging aspect of the puzzle is design—engineering entirely new proteins that don't exist in nature to hit specific targets.

The ESM Atlas

Biohub has now cracked the design problem with their groundbreaking system, ESM Atlas. This system is a world model of protein biology, capable of mapping, predicting, and designing proteins on a scale never seen before. The ESM Atlas covers 6.8 billion protein sequences and 1.1 billion predicted structures. This vast dataset is the result of training a language model, ESMC, on 2.8 billion protein sequences from across all of life. ESMC didn't just learn what proteins look like; it grasped the fundamental "grammar" of biology.

The process

ESM Fold 2, a component of the ESM Atlas, takes the representations generated by ESMC and outputs atomically resolved 3D structures. This process uses a looped transformer architecture that scales efficiently at inference time. Researchers have already used this system to design protein binders against five cancer and immunology targets.

The computational search for these structures, which would have taken billions of years to generate experimentally, was completed in weeks. This staggering acceleration highlights the power of AI in accelerating scientific discovery. All three models—ESM Atlas, ESMC, and ESM Fold 2—are fully open source, allowing researchers worldwide to leverage this technology.

Practical tips for leveraging AI in protein design

Accessing the tools

To dive into the world of AI-accelerated protein design, start by exploring the resources available on Biohub's GitHub page. The open-source nature of the ESM Atlas, ESMC, and ESM Fold 2 models makes them accessible to researchers and enthusiasts alike.

Understanding the models

Familiarize yourself with the ESM language model, which has been trained on a vast dataset of protein sequences. Understanding how ESMC learns the "grammar" of biology can provide insights into how to harness its power for specific research questions. ESM Fold 2, with its looped transformer architecture, is particularly powerful for generating atomically resolved 3D structures.

Applying the technology

Consider how these tools can be applied to your research. Whether you're working in cancer treatment, immunology, or another field, the ability to design proteins with precision can open new avenues for discovery. Think about specific targets you might want to hit and how the ESM Atlas can help you design proteins to achieve that goal.

Important takeaways

The power of AI

AI has the potential to revolutionize fields like protein biology by accelerating the discovery process. The ability to map, predict, and design proteins at an unprecedented scale is a game-changer.

The role of open-source technology

The open-source nature of the ESM Atlas, ESMC, and ESM Fold 2 models democratizes access to this powerful technology. Researchers worldwide can leverage these tools to advance their work, fostering a collaborative and innovative scientific community.

The future of protein design

As AI continues to evolve, the future of protein design looks brighter than ever. The ability to engineer proteins with precision can lead to breakthroughs in cancer treatment, immunology, and beyond. By understanding and harnessing the power of these tools, researchers can push the boundaries of what's possible in protein biology.

Conclusion

AI-accelerated protein design is not just a futuristic concept—it's a reality that is already transforming fields like cancer treatment and immunology. By leveraging the power of the ESM Atlas, ESMC, and ESM Fold 2, researchers can design proteins with unprecedented precision, opening new avenues for discovery and innovation. As these tools become more accessible and powerful, the future of protein biology looks brighter than ever.