The Piano Lab

I recently proposed an open source project. Others might useful.

Question 1
Describe the problem you expect your residency to focus on and your proposed solution or direction for identifying a solution. Aim for a level of explanation that a smart college freshman could understand.
--
One reason Bonolo and I started Drizzle Health was because we couldn't figure out why Tuberculosis was killing millions of people every year, when we had PCR.

We started by wanting to invent a new test. Then we realized the problem in global health isn't just new, shiny, state of the art technology. It was cost and distribution. Every donor's posterchild, PCR tests were fundamentally too expensive at $10 each, even after subsidy, to ever reach most patients. For TB, we decided to build MagnaSlide, which made samples 10x better, which increased widely used microscopy's sensitivity from 50% to 95% for less than a dollar.

But it did get me wondering; why did PCR become this dominant universal solution? My conspiracy theory is as follows: PCR exploded right as computational power became widely available in the 80s-90s (and later in 2000s when computation costs REALLY started to plummet with Moore's law). With nothing available, funders funded more and more implementation around the globe (without understanding the limitations of hub-and-spoke outside the West, smh). This created a self-reinforcing cycle: genomic advances fed computational advances, which enabled more genomic advances. The Human Genome Project, massive open-source software development, plummeting sequencing costs—genomics benefited from catalytic moments that phenotypic approaches never experienced.

During the same period, the newer generation of scientists were trained on genomics in the West, and so now everyone just assumes that and wants that.

They just didn't have to do more with less. PCR is great, don't get me wrong. But using it everywhere, including our imagination, has meant we're also trapping the world into the limitation of PCR and its costs, not just its strengths.

Think about where the IVD industry is placed in the care cycle. You get sick, doctor orders a test and you get results. Why can't we test air at airports so you don't get sick in the first place? For applications like continual air sampling in an airport, we need a $0.01 test, not even a $10 test after subsidy as NAAT claims for PCR. These new models of preventative diagnostics (or Large Scale Diagnostics as we're calling them) are fundamentally not possible with genomics because of reagent and other costs.

Last year, we set out building a true point of care device for community workers in resource limited settings (since they had the most contact with presumptive patients) - We were using Mueller matrix polarimetry—an optical technique that measures how bacteria scatter polarized light, creating unique "fingerprints" that can identify pathogens in conjunction with MagnaSlide. The off-the shelf prototype cost under $500 with near-zero per-test costs after construction. No expensive reagents, no amplification steps, no complex infrastructure. We'd take the data, cross reference it with a library of phenotypic signatures form bacteria and return a result.

Except such a library hardly exists. Unlike genomics, which has vast standardized databases, validation frameworks, and analytical tools, there was literally nothing comparable for optical properties of pathogens. This wasn't just a technical gap—the concurrent evolution of genomics/omics with computation and fields such as bioinformatics hadn't happened for phenotypic data.

(Besides, genomic data is overkill for such screening. What we should be looking for is "what is the least amount of data we need to make a decision." Lowest fidelity data that helps make (clinical) decision, not necessarily high fidelity data for everything.)

That's why I want us to build Open BOPD (Open Bacterial Optical Properties Database)—the world's first standardized repository for bacterial identification using Mueller matrix polarimetry (to start with). It's not just a database; it's the
missing infrastructure for an entire alternative diagnostic paradigm. By creating this foundation, Open BOPD will enable truly affordable diagnostics for applications where genomic approaches will never be economically viable, no matter how economies of scaled. This will have all kinds of bacterial phenotypic properties; recent research has even shown AMR correlated features and phenotypic observations. The point is not our device will work better, but that a new class of devices can now emerge.

Question 2
Why did you pick this problem to work on? What’s novel about it?

--
The genomic/omic revolution that transformed Western medicine has been falsely assumed to be the universal solution for all diagnostic contexts (it is well known that centralized diagnostic systems hardly work in low and middle income countries) - PCR is just an easy answer to fund and push.

PCR's invention coincided with the computational revolution of the 1980s, creating a self-reinforcing ecosystem where each genomic advance made the next more valuable. This historical divergence explains our current predicament.

Phenotypic approaches never experienced comparable catalytic moments or infrastructure development.

While genomics benefited from landmark initiatives like the Human Genome Project, ENCODE, and massive open-source software development, phenotypic methods remained fragmented across academic labs without the standardization, repositories, and tools that transformed genomics from technique to paradigm. This historical asymmetry has arguably (in my experience) trained a generation of scientists to think "genomics first" without questioning whether it's always the appropriate approach. Scientists trained during this period proliferated the genomic mindset globally, establishing it as the unquestioned standard—despite being fundamentally misaligned with needs in resource-constrained settings. It's one shoe fits all for the world (I wrote about in my Drizzle Health newsletter, examining how USAID's approach to global diagnostics created dependency rather than sustainability and now that it is gone....)

The novelty of Open BOPD lies in recognizing and correcting this historical imbalance by building the infrastructure for phenotypic approaches that genomics has enjoyed for decades and enabling a new generation of ultra-affordable diagnostics that provide just enough data for effective decision-making. Because light microscopy (and newer approaches to it) are so cheap, and AI-imaging wants to democratize microscopy and save everyone, now is a good time where a shared open database can really help development of fundamentally 1 cent test techniques.