How AI is Supercharging Biosecurity

AI is unlocking a new era of biological innovation. Coupled with our growing ability to read DNA using next-generation sequencing, AI tools can help us more deeply understand and more efficiently engineer biology. Ginkgo Bioworks is at the forefront of using AI to solve modern problems with biological solutions.

This new era brings great opportunities, but also great risks. AI tools are making biology easier to engineer for everyone, including malicious actors. At Ginkgo Biosecurity, we recognize that biosecurity is critical to counter these risks. By utilizing AI tools to detect, characterize, and respond to biological threats – both engineered and natural – we’re working to ensure that the convergence of AI and biology is not just innovative, but safe. We see AI as an integral part of the solution to countering threats both real and potential.

The Future of AI and Biosecurity

In August, Ginkgo Bioworks and Google Cloud announced a five-year strategic partnership intended to enable Ginkgo to develop and deploy AI tools. Ginkgo has massive, high-quality biological datasets and we are regularly adding new environmental data to our databases. Sequencing analysis can be incredibly resource intensive, making AI an invaluable tool that helps our bioinformatics teams do their work more efficiently. 

For example, when we collect environmental samples through our global bioradar network of monitoring nodes (primarily at airports), we turn physical biology into data, which can be read by an AI, making it faster and easier to spot anomalies when they emerge. The partnership with Google Cloud can help us speed up our AI development so we can further build our capabilities to detect biological threats.

There are many areas of opportunity this partnership could enable. For example, further development in AI-driven metagenomic detection may provide early alerts to anomalous samples. AI could also assess the magnitude of mutations in genetic variants of pathogens to anticipate their susceptibility to vaccines. By partnering with Google, Ginkgo is poised to take a leap forward in the management of biological risks. We detailed the way we see the opportunities recently at the AI Fringe event in London. 

Engineered Threat Detection

Monitoring known pathogens is the backbone of our biosecurity platform, but it’s not the whole story. The ability to detect unknown pathogens is an essential part of a truly effective biosecurity system. That’s where Engineered Nucleotide Detection and Ranking (ENDAR) comes in.

ENDAR is an AI tool developed in partnership with the US intelligence community (IARPA)  to detect and identify genetically engineered biology. It can classify sequencing output as natural or engineered by looking at the common signatures and characteristics of genetic engineering, and create prioritized lists of findings for analysts to investigate. It does this by comparing a sample genome to known pathogens, flagging indications of engineering, and quantifying the number of suspicious or unlikely differences between a given region of the sample and a known version of the pathogen. This technology can help determine where and how a sample was engineered, where changes were made in the microbial genome, and where any novel genes came from, eventually leading us to be able to pinpoint attribution of an engineered sample of interest.

The ability to detect, characterize, and analyze engineered biological threats is part of a set of tools for deterrence against malicious actors that might attempt biological attacks. ENDAR can also help national security and biodefense leaders to identify engineered threats and mitigate risks associated with them.

Epidemiology and Risk Modeling

Ginkgo is unique because we generate such an incredible volume and quality of data. Thankfully, sifting through mountains of data is in AI’s wheelhouse, meaning biosecurity is a field where AI assistance is a natural fit.

Ginkgo uses various AI tools to empower different aspects of our epidemiology and risk modeling analysis work. Bayesian machine learning powers our forecasting and simulation models, which can also assess intervention strategies; gradient-boosted regression allows us to predict geographic areas that are at high risk for outbreaks; natural language processing helps us analyze media sentiment data and travel warnings; and anomaly detection means we can flag deviations from normal patterns of disease occurrence and transmission.

A Responsible Approach to AI

AI represents great potential for breakthroughs, but it also comes with risks. Ginkgo takes these risks seriously in the design of our platform, which undergoes rigorous in-house testing before deployment to ensure that it is secure. Our goal is to deliver biosecurity protection as an intrinsic part of Ginkgo’s offerings, similar to how users can expect some level of cybersecurity when using leading digital tech platforms. 

It’s important that entities doing testing and evaluation research on this topic are incredibly careful about sharing the lessons they learn—so that defensive patching, tools, and regulation can be built before bad actors identify opportunities to cause harm, should they arise.

AI tools could substantially enhance the public health response to any kind of emerging disease, natural or nefarious. The key feature of AI for this application is the ability to learn patterns and to identify anomalies—the signals in the data that allow rapid response to emerging threats. With robust biosecurity infrastructure, we’ll be able to perform the regular, systematic and secure monitoring that, combined with better AI tools, can help enable national security and public health actors to make informed choices.