Comments
The Nucleic Acid Observatory (NAO) aims to improve humanity's ability to detect new pathogens, especially engineered ones. We are evaluating disease surveillance methods, developing pathogen detection algorithms, and running a pilot biosurveillance system. Additional funding would allow us to expand our research into detection methods. You can donate on our website.
Among the most dangerous biorisk scenarios is that of a 'stealth' pathogen, spreading without distinctive symptoms[1] and only later causing severe health impacts.[2] Such a pathogen could cause enormous harm, as a very large fraction of humanity could be infected before anyone notices. The world today is not equipped to detect a stealth pathogen before it's too late; however, improvements in monitoring technology could enable dramatically earlier detection and response, substantially reducing the harm caused.
While the 'stealth' scenario is one where we think the benefits of early detection are strongest, it is not the only one where our work is valuable. We discuss other benefits below.
If a new pandemic were spreading, how would we know? At a high level, we'd need to get samples that might contain the pathogen, learn what the samples contain, and analyze that data for pathogen signatures. We're working to improve the state of the art in all three areas:
We currently work with two sample types:
We've also put substantial effort into evaluating other sampling options, such as indoor air, blood, and pooled airplane lavatory waste, and hope to expand the range of sample types we work with in the future.
After collecting the samples we need to extract the nucleic acids. Among the many potential extraction methods, we've identified ones that optimize for viruses. These methods select for properties common to viruses, such as being smaller than human or bacterial cells. After extraction we use metagenomic sequencing to learn what raw nucleic acid sequences our samples contain.
Sequencing data needs substantial analysis and interpretation to be useful, and identifying novel or engineered pathogens is a serious difficult research problem. We've identified a wide range of data signatures that could indicate the presence of a new pathogen and are exploring how to apply them to the detection problem. Some of these we've developed into tools we can apply today, others we expect to have ready in the next few quarters, while others remain substantial research projects where we're collaborating with outside academics.
We expect work in this area to be our largest effort over the next few quarters.
We would like to hire additional staff in a few areas:
Computational: our current best detection method is an important step forward, but it's far from solving the whole problem. There are many more potential strategies for computational threat detection than we've been able to explore, and another junior or senior researcher would allow us to make faster progress here. This work is very parallelizable, where there are several independent lines of exploration we'd like to pursue.
Wet lab: we recently opened our own wet lab, have hired a second wet lab research scientist, and don't intend to hire additional wet lab staff in the next few months. Depending on how our methods development and pilot biosurveillance work goes, however, by mid 2025 we might want to invest in hiring a wet lab technician to free up our research scientists for work where they're uniquely capable.
We are running a pilot biosurveillance system that we estimate would flag blatantly engineered viral threats today. At our current scale, however, it would likely fail to do so before the majority of people had been infected (simulator). We would like to scale this system to, and ideally beyond, where it could flag such a pathogen before 1% of people had been infected.[3] Piloting the system has also been giving us valuable operational experience, illuminating the real-world problems that come up in building and running a biosurveillance operation; a larger scale pilot, especially of swab sequencing, would bring new challenges and insights. We don't want to end up in a position where we have a good model of how a biosurveillance system would work in theory if only someone would build one, and then building one turns out to be much harder or generate very different data than our modeling assumed.
On the other hand, because of our grant structure we think our pilot detection work is especially susceptible to trading off against funding from Open Philanthropy. As discussed below, we do not anticipate applying funds we raise here to our pilot unless the donor specifically requests.
A final way additional funding would help would be by giving us the operational flexibility to test out new approaches without needing to raise funds specifically for that purpose. For example, we were able to start work on our swab sampling effort relatively quickly once we realized how promising it was because we had some flexible funding on hand. If other opportunities come up we would like to be in a position to move quickly on them.
In general, we hope that donors will trust us to consider these and other options and spend NAO funds in the way that most improves our ability to detect stealth pathogens. On the other hand, some of the tradeoffs here are not about our core area of expertise but instead about the broader world: how likely is it that someone tries to create a pandemic like this in the next year vs farther in the future? How much of the risk comes from very clearly engineered things that are easier to detect vs more sophisticated attempts that put substantial effort into hiding what they're doing? If you're interested in contributing a substantial amount and have strong views on these questions, please reach out to [email protected]: we have many efforts that are primarily funding-limited and are open to considering your priorities in our prioritization.
The best way to get an overview of our recent work is to read our recent blog posts, and especially our Fall 2024 Updates post. These give the most detail about where we're putting our work and how that's been going.
The NAO began as a collaboration between the research non-profit SecureBio and Kevin Esvelt's Sculpting Evolution group at MIT. We are in the process of transitioning the project to be entirely under SecureBio, and we expect to complete this by the end of 2024.
The NAO is structured as two sections, a Near-Term Detection project to build, operate, and improve a pilot early warning system, and a Robust Detection project to develop methods that can identify threats beyond the capabilities of our current systems.
There are nine people working full-time on the NAO. On Near-Term Detection we have one lead, one computational research scientist, two wet-lab research scientists, one research analyst, and one research assistant. On Robust Detection we have one lead and two computational research scientists.
If we continue at our current level of operations, in 2025 we will spend $2.2M:
| Expense | Amount |
| Staff | $1.3M |
| Pilot Detection System | $520k |
| SecureBio Overhead | $125k |
| Wet lab consumables | $120k |
| Compute | $100k |
| Lab space | $85k |
| Office Space | — |
Our office space is generously covered by Open Philanthropy.
The NAO currently has $2.2M available in funds and firm commitments, for a runway of 12 months.
A successful NAO would create a system able to detect stealth pathogens early enough in their spread that there's time to avert the worst outcomes. This has additional benefits through deterrence: the best way to stop someone from creating a stealth pandemic is a detection system that reduces their chances of success to where it's no longer worth their effort and risk.
There are also benefits in the case of non-stealth pathogens, because a system based on untargeted sequencing can be very rapidly applied to any new pathogen once the genome is known. For example, if we had been operating this kind of system in January 2020 when the SARS-CoV-2 genome was first shared we could have immediately checked whether any of our samples had ever had a match. We could then have continued using our system to track that genome until a more sensitive targeted system was in place.
There are also a range of other benefits:
The biggest reason why we expect people might not want to fund the NAO is that they don't think that stealth pandemics are a sufficiently likely or pressing problem. We also don't have a reliable estimate of the likelihood: on one hand humanity hasn't ever had an engineered pandemic, let alone an engineered stealth pandemic, but on the other creating pathogens that could potentially cause this kind of harm has only recently become plausible.[4] Given this trend, the scale of the devastation this sort of pathogen could have, and how few resources are currently devoted to this threat, however, we think this is among the world's most pressing problems.
If we detect something and raise the alarm, humanity might not respond effectively or it might be too late for effective response. An alarm is only valuable to the extent that it causes people to usefully change their behavior, and it could be hard to get people to take claims of a stealth pathogen seriously. We are also concerned about this, and are working with a range of stakeholders to develop plans for validating warnings and taking appropriate downstream actions. Still, we think there's a lot more work that could be done here and we would be enthusiastic to work with others on how they could help us turn detection into action.
As with many projects that have received substantial funding from Open Philanthropy, it’s reasonable to be concerned about your donation funging against their support: you may be less interested in funding us if Open Philanthropy then counterfactually decreases their funding by a similar amount. We agree this is an important consideration.
Our current funding from Open Philanthropy comes in two main forms: a large grant for general support intended to cover several years and a series of smaller grants specifically to support sample collection and sequencing. We don’t expect more general support from Open Philanthropy for at least the next year, and so expect non-Open-Philanthropy donations to make a significant counterfactual difference to our ability to grow and support our team over the next one to two years.
We’re significantly more uncertain about the counterfactual impact of donations to support sample collection and sequencing: Open Philanthropy has so far been willing to give additional dedicated grants. Our current grant covers our work through approximately the end of February, we plan to ask for additional funds in early December, and we expect to hear back sometime in January. We don't know how likely they are to provide continued funding. Non-Open Philanthropy funding here is therefore quite high variance: it could play a key role in allowing us to expand and continue our surveillance work, or it could entirely funge against Open Philanthropy's support. Given this uncertainty, we are currently not raising funds for our pilot biosurveillance work. If you're specifically interested in funding this work, however, we encourage you to get in touch.
Even if funding from this appeal funges against Open Philanthropy more than we expect, we still see substantial value in diversifying our funding pool. This would give us more stability and more independence from Open Philanthropy, which is good for us and others who believe in our mission and goals.
If you'd like to support the NAO you can make a donation to SecureBio earmarked for the NAO project on our website. Our donation form supports credit cards, debit cards, and, more efficiently, bank transfers. Donations from US residents are tax-deductible. To explore making a tax-deductible donation from the UK or EU, or to discuss alternative ways to give, please write to [email protected].
This could be something spreading with no symptoms, or with symptoms that are generic enough that no one notices anything unusual.
HIV is a good example of a pathogen with this pattern, though it spreads relatively slowly. An engineered stealth pathogen could combine rapid spread with a long asymptomatic period.
We estimate that expanding to weekly wastewater sequencing would, in the median case, allow detection of a viral pathogen that sheds like influenza or SARS-CoV-2 before 1-3% of people in the monitored sewersheds had ever been infected (simulator). This would double our annual wastewater sequencing costs, from $470k to $940k. Similarly, our swab sequencing efforts are very promising. While this portion of our pilot is at a much earlier stage and there are a lot of unknowns, we estimate that at a given cost it would be about five times more sensitive in flagging a respiratory pathogen than wastewater sequencing (simulator). We would be enthusiastic to expand this work to daily sampling, for a marginal expense of about $100k annually.
Appendix 2 of Securing Civilization Against Catastrophic Pandemics attempts to estimate this probability, and gives an annual risk of 0.28% to 13% once credible blueprints for a stealth pathogen are publicly available.
