The Piano Lab

Environmental Bio-surveillance

A key consensus ported over from the Life Sciences into biosecurity is that of accurate testing. Which is important. This consensus is atleast incomplete.
The focus on making tests faster, more sensitive, higher fidelity has lost the question whether accuracy is the right constraint. Perfect information about 0.01% of samples (or environment) is still very limited when it comes to environmental bio-surveillance.

Two insights:

/

Biosecurity is a coordination problem
Biosecurity is a coordination problem, as much as it is detection/technology problem. You’ll have a hard time monitoring what you can’t access. And you will have a hard time accessing what doesn’t benefit the owner of that access.
Consider that about two-thirds of land in the US is owned privately. More than 90% of agricultural land. Or more globally, small scale farmers (<2 acre) represent about 70% farms in India, Vietnam, etc. They live on <$5 a day. Why is this important?
Because, in effect, these private farmland owners control the majority of zoonotic spillover interfaces. The FAO estimates 60% of all human infectious diseases are zoonotic in origin, with 75% of emerging diseases involving species jumping. Emerging infections still mostly originate not from labs with policy mandates, but from interfaces like livestock interacting with wildlife.

If solutions provide no incentive for farms to participate (and on the contrary cost them money) the unrealistic burden of critical coverage will fall on technology invention alone (ie we’ll need better and better technology to compensate for bad samples**) We need innovation in social business models as much as in technology because unless our plans help these people make money, they will never participate in surveillance.
We’re asking farms to voluntarily report problems that could trigger culling and operational shutdowns, affecting their finances and livelihoods. That private actors will not bear costs is abundantly clear but that’s because their incentive structure is broken: they’re incentivized to delay reporting and hide problems.

Biosecurity as a public good provided by governments is very limited in what it can currently achieve. Donors need to start thinking about biosecurity infrastructure that also benefits private actors financially. Can it reduce livestock mortality? Lower insurance premiums? Enable premium market prices? If you’re telling test developers to innovate while adoption remains impossible without hub-and-spoke models, there’s a structural problem. There needs to be public-private partnership.

/

The Technology Trap
The general sense in biosecurity is that faster, more sensitive tests, sequencing data is better. This is carryover from Life Sciences. Take PCR (NAAT) for example, which through volume guarantees is now offered at around $8 for TB (after subsidies). How much sample does it process? 2ml. So in essence, a NAAT machine costs $4/ml. This works nicely in life science where human samples are small. The problem is that when you apply this to environmental surveillance, where volumes are in the thousands and millions of liters, testing even a single lot of lettuce in the food supply chain becomes untenable. (And most samples/distribution in nature are heterogenous).
I’d argue PCR’s dominance isn’t just technical but institutional. Major donors with the right intentions spent hundreds of millions scaling PCR platforms globally. Equipment manufacturers built business models on $10-50 reagent cartridges. Academic careers were built on genomic methods. An entire generation of scientists was trained to think sequence first, ask questions later.

But IVD as an industry was not designed with environmental surveillance (or biosecurity) in mind, and the use of its tools make bio surveillance economically impossible. The volumes, fidelity, timeframes, and incentives are different. Through interviews, we’ve learned for example, that most food lots are not even tested before they’re shipped. Lettuce, for example, moves through in 3-15 hours, and a PCR takes 6-24 hours plus enrichment time (upto 2 days). By the time you have results, the product is already at Chipotle. There are 48 million foodborne illness cases in the US yearly. Testing costs time and money. And runs the risk of shutdown.

There are fundamental cost structure limitations imposed by PCR. Which in-turn limit our imagination about how biosecurity should be done.
We have a testing post facto model, for example, instead of preventative models. Why don’t we have a pathogen air quality monitor? We can get faster at finding TB patients but we can’t prevent transmission because we can’t test transmission at scale. Not with PCR. Continuous air sampling at airports would cost tens of thousands of dollars daily per location at current test prices.

We’re stuck in a high-cost paradigm fundamentally incompatible with prevention. We have to rethink the constraints. If you want continuous air sampling at airports in all kinds of places to catch the next pandemic early, you don’t need a $20 test. You need a $0.01 test. (And a $0.01 test will likely be borne out of the minimum information required to make a decision. Not through racing for the highest fidelity data, which may not be required to make clinical decisions.)***

Early warning systems need to solve for size, not just fidelity.

/

A Solution:
One solution that me and my team at Drizzle have been pursuing, is Large Scale Diagnostics. How do we sample thousands of liters of samples, concentrate organisms of interest, and turn them into 1ml samples to use with existing diagnostic facilities public and private partners. Our MagnaFlow line of product, a feat in polymer and millifluidic design and engineering, runs all the washwater from a lot at the processing plant, and captures specific bacteria of interest. This can then be tested using existing tests available on site.

This way producers can know for sure if the lot they’re sending out is contaminated or not. It reduces recall risks and costs, and further reduces labor costs and insurance premiums, directly affecting financial bottom lines for each processer. In effect, the innovation isn’t just the technology, but also that the product creates its own adoption incentive.
Products with the same framework can extend globally, including for the Indian farmer living on $5/day. The biosurveillance network will emerge as a side effect of giving millions of private actors tools that help them make money.

/

Biosecurity requires two fundamental shifts. First, in incentive design: stop treating surveillance as a public good. Two, in technology constraints: stop optimizing for clinical sensitivity; start optimizing for surveillance coverage at economically sustainable costs.
The next pandemic won’t be stopped by better PCR. It will be stopped by redesigning the economic and technological foundations of biosurveillance itself.

—

**

One big realization for us while working with TB, was the bio-stack. Garbage In, Garbage Out. A lot of work today is being done for better technologies at the detection layer and the data layer. If the sample is bad (as is the case in TB), detection mechanisms have be very sensitive to detect organisms. Very sensitive = higher costs/test. And this is true in other diseases too. Like Malaria. Or until a couple of decades ago, even pap smears.

Drizzle’s platform works at the sample and prep layer, with easy integration into existing detection.

By simply making the sample 10x better for 10 cents, we’ve been able to make TB diagnostics 60% higher performance.

/

***

Part of it also the traditional business model for dual-use technologies such as those used for biosecurity. There’s a lot more clinical native technologies that’re adopted for biosecurity, than the other way round.