Synthetic Biology: A Quarter-Century of Engineering Expectations

This year marks the 25th anniversary of synthetic biology, dating back to two groundbreaking papers published in Nature in 2000. Before this milestone, biology was primarily in the science domain—researchers observing, describing, and analyzing natural phenomena.

What the authors conceptually did differently was to apply engineering principles to living systems (bacteria), constructing a genetic toggle switch and an oscillator - essentially biological equivalents of an on/off switch and a clock. This sparked hope that cells could be engineered with precision, with bioengineers building genetic circuits for specific functions.

This evolution intuitively seemed like the logical next phase for biology,

Bioscience → Bioengineering,

which followed an innovation pattern we've seen in other domains:

• From analytic chemistry evolved chemical synthesis, which revolutionized the development of pharmaceuticals and materials

• Aerodynamics laid the knowledge foundation for aerospace engineering

• Computer science gave rise to software engineering

Yet a quarter-century into this journey, despite notable successes, we must acknowledge that biology hasn't fully yielded to pure engineering approaches.

The Challenge of Complexity

Taking a purely mechanistic engineering approach to biology meant ignoring its inherent complexity. This has led to oversimplified "proof of concept" systems that function in an undercomplex lab setting but break when implemented in industrial contexts.

One phenomenon makes engineered biological systems particularly challenging: context-dependency

• The same gene might make lots of protein or none at all, depending on where it is integrated into the genome

• A genetic circuit functioning perfectly in E. coli might be completely silent when transferred to another bacterium

• A microbial strain that performs well in a laboratory flask may behave unpredictably when its environment is changed to a commercial-scale bioreactor

In biological systems, every component connects to numerous others. Sometimes the connections are known, sometimes not. A seemingly minor modification in construction can cascade through the system and break non-obvious emergent properties.

Expanding the Framework: Biodesign

Should we abandon efforts to make biology engineerable and accept the limited scope of products that came from bioengineering? If you believe, like me, that biology is the most sophisticated technology on our planet, giving up on it would seem like a massive lost opportunity.

But if the direct path from bioscience to bioengineering has proven challenging, perhaps we can expand our framework to the design domain:

Bioscience → Bioengineering → Biodesign

Just like the innovation pattern that science gives rise to engineering, conventional wisdom holds that we must master the engineering domain before we can move into the design domain of a given discipline.

To learn more about how engineering relates to design and the other two "domains of creative exploration", science and art, I suggest reading Neri Oxman's essay "Age Of Entanglement"

To be clear: I'm not advocating building high-end "designer bioproducts". I'm talking about design as an approach that focuses on users and usability.

Examples:

• Modern manufacturing was initially driven by fabrication methods and then evolved into product design that considers the entire lifecycle, from materials sourcing to end-of-life recyclability.

• Car manufacturing progressed from mechanical engineering (making vehicles that function reliably) to automotive design that integrates performance, ergonomics and sustainability.

• The internet was a technical experiment, powered by protocols built by engineers. The World Wide Web was later added as a design layer that made it widely accessible.

But if the past suggests that bioengineering must precede biodesign, what should make us confident that we can take a shortcut this time? The answer is, as so often these days, artificial intelligence.

Leapfrogging the Engineering Domain

AI systems excel precisely where pure engineering approaches struggle - leveraging context and complexity instead of stripping it away. Contextual embeddings allow advanced models to take "surrounding" information into account, to better interpret or capture the meaning of a piece of information.

This holistic approach allowed AlphaGo Zero to liberate entirely from human Go strategies (which tend to focus on "local battles" instead of taking the entire board into account). With nothing more than the knowledge of the fundamental rules of the game, AlphaGo Zero reached the strongest level of play in history, using strategies that no human came up with in millennia.

For biology that means that, conceptually, instead of engineering predictable biosystems from the ground up, biodesigners could create holistic solutions that account for context.

Bioscience → Biodesign shortcut?

AI models can enable this by identifying emergent properties within biological data sets, by recognizing relationships between genetic sequences, environmental conditions, or cellular behaviors.

We can see the first signs of this biodesign paradigm in drug discovery: AI models are now helping medicinal chemists design small molecule drugs and predict their pharmacokinetic properties in real-time. These capabilities require seeing the drug in the context of the human body -- how it is absorbed, distributed, metabolized, and excreted -- and they play a crucial role in determining success in a clinical trial.

The Biodesign Opportunity

I expect similar approaches to quickly find their way into other sectors. I envision a near-future where biodesigners specify desired outcomes - such as organisms that produce specific compounds - and AI tools suggest entire novel genomes for both optimized performance and commercial viability.

This design-first approach would encompass:

• Predicting how engineered organisms will behave during scale-up

• Suggesting optimized cultivation conditions and feedstock requirements, based on minimal input data

• Tweaking extraction and purification methods, based on all sorts of abstract data, including location

The result would be fewer engineering iteration cycles, faster development timelines, and biological solutions more aligned with market needs, with competitive unit economics.

The Expanded SynBio Framework

To be clear, AI isn't a magical solution that will instantly unlock the trillion-dollar bioeconomy. We still need advances in genetic engineering tools and measurement technologies (data!), as well as infrastructure and more favorable policy making.

From an investor point of view, what makes me bullish is that "AI-native" synbio companies (e.g. Cradle, Cambrium, Fungtional) already exist and prove to be successful across different markets and business models.

As synthetic biology enters its second quarter-century, complementing engineering approaches with AI design could help us fulfill the field's original promise: harnessing biology's power to address our most pressing challenges in health, materials, energy, and sustainability.