AI-Designed Proteins: Two decades ago, the notion of engineering designer proteins was merely a distant dream. Fast forward to the present, and thanks to the integration of artificial intelligence (AI), customized proteins have become a common occurrence. These tailor-made proteins possess specific shapes or components, granting them novel abilities beyond what nature offers. This technological advancement holds immense promise, from creating longer-lasting medications and protein-based vaccines to developing eco-friendly biofuels and proteins capable of digesting plastics.
The realm of customized protein design heavily relies on sophisticated deep-learning techniques. Large language models, such as the AI driving OpenAI’s renowned ChatGPT, play a pivotal role in envisioning millions of protein structures that transcend human imagination. This results in the rapid expansion of the library housing bioactive designer proteins.
The empowerment derived from this progress is undeniable, as expressed by Dr. Neil King from the University of Washington, who emphasized the transformation of once-impossible tasks into routine endeavors within the past year and a half.
However, as the use of newly designed proteins gains momentum in fields like medicine and bioengineering, scientists are confronted with a pressing question: What safeguards are in place to prevent the misuse of these technologies?
In a recent essay published in Science, two experts in the field, Dr. David Baker, Director of the Institute for Protein Design at the University of Washington, and Dr. George Church from Harvard Medical School, delve into the necessity of implementing biosecurity measures for designer proteins. Similar to ongoing discussions about AI safety, the authors propose the need to address biosecurity risks and formulate policies to ensure that custom proteins do not pose unintended threats.
The essay advocates for embedding barcodes into the genetic sequences of synthetic proteins. In the event that a designer protein becomes a potential hazard, such as triggering a harmful outbreak, its barcode would serve as a traceable marker, facilitating the identification of its origin. This system essentially provides an “audit trail,” adding an extra layer of accountability to the development and use of custom proteins.
The convergence of designer proteins with AI necessitates a parallel consideration of biosecurity policies. Over a decade ago, Baker’s lab utilized software to create a protein named Top7, demonstrating the ability to explore uncharted territories within the protein universe. The advent of AI-accelerated protein design, with techniques like structure-based AI and language models like ChatGPT rapidly generated protein structures suitable for real-world applications.
Despite the potential benefits, concerns have emerged about the safety and ethical implications of AI-enabled protein design. Governments worldwide have initiated plans to oversee AI safety, focusing on privacy laws, economic impact, public health, and national defense. While synthetic proteins were not explicitly addressed in initial legislation, new AI regulations are underway, with the United Nations advisory body on AI expected to release guidelines on international regulation in the coming year.
To navigate potential regulatory challenges, the authors emphasize the importance of documenting each new protein’s underlying DNA. Building a comprehensive database of synthetic DNA sequences could aid in identifying potential risks, such as proteins resembling known pathogenic structures. However, the authors also acknowledge the need to balance data sharing with protecting trade secrets.
One proposed solution involves integrating safety measures directly into the protein synthesis process. The authors suggest incorporating barcodes, composed of random DNA letters, into the genetic sequence of each new protein. During the synthesis process, a machine would search for the specific code before initiating protein construction. This approach grants the original designers control over who has access to the synthesis process, addressing concerns about sharing potentially dangerous designs.
Implementing a barcode system tied to synthesis machines not only enhances security but also acts as a deterrent against illicit activities, making it challenging to replicate potentially harmful products. The authors envision this strategy as a global effort, requiring collaboration among scientists, research institutions, and governments to ensure the safety of designer proteins on a worldwide scale.
While the road ahead may be challenging, the authors remain optimistic that open discussions about biosecurity will not hinder the progress of the field. On the contrary, they believe such discussions can foster collaboration across different sectors and engage the public in a dialogue that promotes the continued advancement of customized protein design in a safe and responsible manner