AI and Synthetic Biology: Balancing Innovation With Biosecurity
Artificial intelligence is revolutionizing biotechnology innovation, but industry stakeholders must work together to mitigate its risks.
The following article is an opinion piece written by Adam Clore. The views and opinions expressed in this article are those of the author and do not necessarily reflect the official position of Technology Networks.
Artificial intelligence (AI) is improving everything, from enabling researchers to extract actionable insights for drug development, to predicting protein structure and function.
Recently, the scientific community has recognized a new potential for synthetic biology: to improve biosecurity. The 2024 Nobel Prize in Chemistry focused on protein design and was shared by the inventors of AlphaFold2, an AI tool from Google DeepMind that can be used to predict the structure of proteins. A commentary co-authored by Microsoft chief scientific officer Eric Horvitz at that time hailed the tool as one of several that could be used to rapidly model emerging pathogens and then design protein-based therapeutics to deploy during a pandemic.
AI-assisted protein design is more efficient than traditional methods of drug development, such as isolating antibodies from patients and synthesizing copies of them. With AI-enabled predictions, drug developers could instantly update antibody-based drugs in response to mutating pathogens. There’s little doubt AI will reduce the overall time and cost of drug development, not only in response to outbreaks, but also for innovating new therapies for a wide range of diseases.
But these innovations also pose risks to biosecurity by opening the door to individuals who may want to misuse AI for nefarious purposes. To counter this, scientists in academia, industry and public agencies must work together to increase the resiliency of biosecurity screening tools –and develop processes for mitigating the possibility that AI will be used to create biological weapons.
Improving biosecurity screening technologies
The main risk posed by new AI tools is that the increased ease of predicting protein structure and function without costly experimentation could make it possible for non-scientists to design novel toxins or modify pathogens to become more harmful. AI-assisted tools could be used to synthesize genes that encode harmful proteins, for example. Nucleic acid synthesis providers can screen orders to try to detect and prevent this activity, but how reliable are these screening tools?
Last year, my company, Integrated DNA Technologies (IDT), was approached by a Microsoft team led by Horvitz to evaluate nucleic acid biosecurity screening tools. We designed a study to assess the ability of these technologies to detect synthetic proteins of concern to the biosecurity community. We engaged multiple nucleic acid synthesis providers, as well as companies that make screening tools.
What we learned was concerning. First, we performed a framing study of two widely used biological screening software tools to see if they could detect synthetic versions of a protein of concern. Both failed to reliably detect altered DNA sequences that encoded the harmful synthetic homologs. That said, we determined that the screening software could be enhanced to improve its surveillance capabilities.
We expanded the assessment to include a wider set of potentially hazardous synthetic proteins and additional screening tools. This time the results varied, with some screening systems outperforming others in their ability to detect proteins of concern. Our conclusion: biosecurity screening systems could be improved through a continuous process of stress-testing them against potential threats, building insights from the results and continuously improving the software to improve resilience against emerging biohazards.
Industry-wide collaboration
Ensuring AI-enabled synthetic biology is safeguarded and not exposed to threat will require coordination between biosecurity screening companies, governmental regulators and nucleic acid providers. To ensure we meet our goal of preventing misuse, while at the same time providing scientists with the tools they need to rapidly develop life-changing advances, we must continue to work together to test and improve biosecurity methods and protocols.
One group that’s helping to advance that goal is the International Gene Synthesis Consortium (IGSC). It provides a screening framework, the Harmonized Screening Protocol, which provides concrete steps gene synthesis companies can take to screen orders and vet customers to lower the risk of misuse.
We all share the responsibility of ensuring the safe and responsible use of AI in the life sciences. AI-enabled technologies can help researchers solve complex problems that were out of reach before and unlock scientific breakthroughs that will revolutionize the treatment of many diseases. By partnering to strengthen our expertise in biosecurity, we can ensure that AI in biotechnology will advance the greater good – and stay out-of-reach of those who intend to misuse it.
