Building Biology from Parts and Services
Complex things get simpler when you can build them from parts. Today, we’re talking about building biology from pieces of DNA or from data services.
Transcript
Looking at a living thing, you don’t necessarily see a complex system composed of interchangeable parts. Why would you? After all, humans didn’t invent biology, biology invented us. Parts are a human concept, and we always want to remember that the possibilities of nature are not limited by our ideas.
But, that said, as far as human ideas go, the idea of parts is a pretty good one. Industrial devices are manufactured from parts. Software is composed of functions, microservices and other modular pieces of code. Essentially anywhere you see a cool piece of technology, it was probably built by connecting and composing simpler devices.
So I don’t think it’s totally arrogant to think that biology might have similar organization, even if it won’t be exactly how humans might build it. And we've seen how applying the concept of parts to biology can be incredibly useful even when the analogy isn't perfect.
For example, if you’re an Austrian Monk in the 19th century studying the characteristics of peas, you might find that heredity gets easier to understand if you think of traits as discrete units that can be passed down without blending.
Yada yada yada, 100 years later, all of modern biotechnology is enabled by our understanding of genes as discrete segments of DNA with a well defined beginning and end.
Around the birth of the modern biotech industry, building biology with parts looked like cutting and pasting segments of DNA. A classic example is the bioproduction of human insulin. Back in 1978, that meant physically cloning a gene from human cells into microbial cells.
In the decades since, a lot of the work of engineering biology has focused on refining our understanding of modular chunks of DNA. We’ve gotten much better at associating specific sequences with specific functions at higher and higher resolution. Where once we knew “this segment of DNA makes something important,” now we might have more precise annotations for what it makes, how much, how stable, how active, in which cells, and so on.
Breakthroughs in DNA synthesis mean that physical cutting and pasting no longer limits how we build. Instead, we can assemble hundreds of genes at a time, selecting specific sequence variations for each gene in any number of nucleotides at any positions we think might be relevant.
The upshot of these innovations is that biological parts, as we once defined them, are increasingly obsolete. We all knew this day was coming. We all knew that it wasn’t strictly true that biological functions could be cleanly mapped to narrowly defined pieces of DNA. It was just a useful simplification that’s become less useful over time. So the time has come to move on.
Progress in the next decade of engineering biology means refounding our understanding of biological parts. Parts are still a good idea. We’re still human beings. We still need simplifications to help us build. But we need these parts to be more expansive and more generalizable to meet the moment of DNA synthesis, data science and AI.
I believe that the new parts are data services. Building biology today is less focused on the boundaries of a piece of DNA and more focused on efficiently generating data to characterize its function.
Biotech R&D manages three different kinds of things. The beginning of a project is a DNA library, a large collection of DNA sequences with targeted and useful variation. The middle is a collection of data services that return measurements of relevant properties for each piece of DNA. And the end of the project is the functional outcome, the most relevant property of all and the goal of the project.
For example, maybe you need an enzyme for green chemical manufacturing. The DNA library would include natural enzymes and targeted variations of them. The data services would include measurements for enzyme activity, specificity, stability, expression, and so on. The functional outcome is your bioproduction costs for purified, active enzyme - the number that goes to the bottom line of profitability.
Or say you’re making a biologic. The DNA library might include designer antibodies, maybe sequences generated by AI. The data services include measurements for binding affinity, selectivity or thermotolerance. The functional outcome is, of course, a clinical endpoint that makes a human being healthier.
The engineer’s role is to build an R&D project from these parts. You’ll compose the library and the right collection of data services to produce your desired functional outcomes. Each choice needs to be carefully considered because the value of the dataset is more than the sum of its parts.
That makes building with data services quite different from high-throughput screening or other data-intensive strategies from earlier eras of biotech. No single sequence needs to be perfect on the first try. No single measurement is likely to directly predict a clinical outcome. Success is an emergent property of the complete dataset, if you've selected data of the right quantity and quality.
Moving from DNA segments to data services is a very different way to think about building biology, but I think it delivers the main thing we need parts to do: simplify the engineering process. Data services can be selected from a menu and purchased from a data factory like Ginkgo, without the need to build out a new experimental workflow in house.
That means you can collect more kinds of data and more data per dollar. It frees up your team to focus on the high level tasks of dataset design and the functional outcome that really matters.
I’m excited for this new era of synthetic biology. I think there’s an opportunity here to fundamentally rethink the way we do R&D. The new data services will need to include some simplifications for the sake of human convenience, just like the old genetic parts did.
But I think we’re getting closer to a match for the realities of biological complexity. And I think that, here at Ginkgo, you can just build with better parts. The history of engineering biology shows that new kinds of biological parts can drive new kinds of biological progress.