Because biology is the result of evolution and not human development, bringing engineering principles to it is guaranteed to fail. Or so goes the argument behind the “Grove fallacy,” first invoked by drug industry observer Derek Lowe in a critique of Intel CEO Andy Grove in 2007. After being diagnosed with prostate cancer, Grove found himself frustrated by what he described as the “lack of real output” in pharma especially as compared to the drive of Moore’s Law in his own industry.
This was a naive and invalid criticism from Silicon Valley outsiders, Lowe argued, because “medical research is different [and harder] than semiconductor research” — and “that’s partly because we didn’t build them. Making the things [like semiconductors] from the ground up is a real advantage when it comes to understanding them, but we started studying life after it had a few billion years head start.” So the very idea of engineering biology by nature is doomed to fail, he further wrote, given that “billions of years of evolutionary tinkering have led to something so complex and so strange that it can make the highest human-designed technology look like something built with sticks.”
But we’ve seen incredible advances in the world of biology and tech the last few years, from AI diagnosing cancer more accurately than humans do, to editing genes with CRISPR. So is it still true that the idea of bringing an engineering mindset to bio is another case of starry-eyed technologist solutionism?
It is absolutely true that we’re very much in the process of discovering biology, still untangling the “technical debt” of evolution. Just when one thinks one understands the biology, another layer of the onion appears. It’s also dangerously easy to break biology, with far greater consequences than broken code — even single-point mutations can lead to disease, and extremely small quantities of certain chemicals can have disastrous side effects. Many of the failures of medicine and especially drug design stem from the complexity and unpredictability of biology.
But the fact that we are still discovering biology doesn’t mean that we can’t design. We can engineer the tools we use to manage biology.
In fact, we have been building and designing tools to control, augment, replace or enhance biology as long as humanity itself has existed — whether it’s taming the jungle to build habitable villages; halting and containing infection; making advanced prosthetics for people who lost their limbs; making synthetic drugs to replace defective parts; or now, even creating functionality that nature never had. We can do this because we empirically learned the properties of those materials, then iterated, designed, and built new structures with them. There is no reason why we should not continue be able to do so for our medicines and bodies.
The only question is how we get there. If discovery is the systematic exploration of ideas and concepts with the goal of understanding the world around us, then design is the mainstay of engineering — where concepts learned in the scientific arena are employed to build everything around us in a repeatable, less time-consuming and more predictable manner.
The way things are right now, we design bridges, but we discover drugs. This is not without cost: Billion-dollar bridges, which we have learned how to design through trial and error, practice and well-tried engineering principles, rarely fail — whereas billion-dollar drug failures are routine, not to mention costly. With design, however, we can plan and progress very systematically along a roadmap and make incremental innovations along the way. Borrowing from engineering, here are principles that allow us to overcome the so-called Grove fallacy and harness biology.
Biology has a hierarchical nature: amino acids are made from atoms; proteins from amino acids; and so on — all coming together to make cells, which make tissue, which leads to organs, which leads to organisms, and then niches, then complete ecosystems. Evolution is the ultimate algorithm. And the ability to facilitate evolution more rapidly to respond better to selective pressure (meta-evolution), has led to mechanisms that reinforce this modularity. A great deal of the hierarchy is known, so if you want to engineer cellular machinery, the parts are proteins; if you want to engineer tissue, the parts are cells; and onward at higher scales.
This isn’t just hypothetical. There are numerous successful examples of this already, from the engineering of myosin (proteins that walk along cell microtubule highways for transport) to CAR-T therapies, where by identifying two key protein modules (the “Legos” in this case) and bringing them together we can engineer patients’ immune cells and thus treat their cancers. Researchers and entrepreneurs used this fundamental aspect of biology to program cells, basically curating from all proteins in the cell a limited set that “plays well with each other” and the rest of the cell to use as a set of Legos for building genetic circuits.
Once we identify the Legos in biology and their properties, we can engineer them and even mix and match them to design novel functionality.
Irreproducibility is a major crisis in modern biology, especially when it comes to publishing papers and being unable to replicate results. But reproducibility is a critical hallmark of an engineering-based approach to biology; it is, by definition, impossible to engineer a process without reproducibility.
One of the principle causes of irreproducibility in biology is the pre-Industrial Revolution, bespoke (literally, hand-crafted) nature of biological experiments even today. This makes most experiments more art than science. But modern technology makes the process of doing biology much more reproducible, from problems of consistency in reagents to re-running and debugging issues. Robotics is one of the most obvious ways, with exact motions now done with the precision of a machine and directed by software.
Machine learning plays a huge part too. The identification of biomarkers (chemical substances we can measure and then target) for disease is currently driven by discovery via a bespoke, one-off process — so the discovery of PSA for prostate cancer, for instance, does not suggest a biomarker for ovarian cancer. Introducing machine learning into the process, however, can turn this handcrafting into assembly-line production. Furthermore, we’re teaching the machine how to fish, allowing not just for reproducibility but for the improvement of accuracy over time, thanks to inputs from additional raw data and identification of complex patterns that humans are incapable of seeing.
A number of companies are already doing this. Apple’s latest watch has been heralded as a the “iPhone moment” where consumer wearables might “morph into medical grade devices.” Entrepreneurs applying deep learning to medicine can use AI/ML and labeled data from generic Apple watch pulse data streams to accurately and precisely identify atrial fibrillation. But more notably — where biology becomes engineering — is the ability to take the very same process they use to detect heart disease and then predict patient illness in many other areas as well (hypertension, sleep apnea and type 2 diabetes).
Such machine-learning driven companies — which can also detect cancer early or identify biomarkers linked to longevity — have all engineered a kind of “factory line.” Given the right input ingredients, they can now mass-produce many tests in a predictable, precise and repeatable manner. It’s yet another way the Grove fallacy is false. The massive advances in computer chips (Moore’s Law), storage (“Kryder’s law”), and genomics — all exponential decreases in cost, 1,000 times over a decade — come merely from 30% improvement year over year. In biology, better reproducibility plus gradual improvement over time plus greater accuracy all add up to even more massive advancements, because even a little goes a very long way.
Testing is the ability to understand exactly where a given product/diagnostic/drug stands, and while the need for testing is obvious, how to test and what metrics to measure success are not. So, the choice and engineering of key performance indicators (KPIs) is critically important here; without this guiding compass, a project could go in the wrong direction.
KPIs are used all the time in engineering, and in all businesses, as a way to define and measure success (or at the very least, progress). But “traditional” biology experiments and drug development haven’t used the concept, since biology was conceptually driven by discovery: how can you assign a KPI when you don’t know what you’ll discover? Now, a new wave of bio startups — drawing on engineering and computer science — are identifying KPIs for measuring molecules synthesized to protein expression, numbers of cells screened, and much more.
The critical part is determining what the right KPIs are, and in engineering biology, there are few precedents, so this can be challenging. But much like in medicine more broadly, the basic principle is: what can be measured can be improved, and those improvements can have huge payoffs. In fact, the evolution from subjective intuition to objective measure is itself another indicator of moving from discovery (biology) to design (engineering), and fits more into Grove’s worldview.
An obvious approach to bring engineering into biology is to apply existing engineering disciplines– materials, chemical, electrical, mechanical, and so on — in the biological realm. Until recently, the ability to quantitatively test and iterate biology was greatly limited. But the rise of numerous, novel quantitative measurements of biology — i.e., big data sets in biology — has opened the door to incorporating other engineering approaches.
For example, there are also companies out there using mechanical engineering principles to bring simulations to predicting the outcomes of surgeries, so that treatments can be engineered, instead of discovering through trial-and-error on patients. By applying the materials-science based engineering technology he learned in solar cell materials design to food, James Rogers used techniques from nanoscience to create nanoscopic barriers that protect fruits and vegetables from spoilage. This isn’t biomimicry, but a way of borrowing the fundamentals of engineering from other disciplines to harness biology.
Engineers at NASA in 1962 could perhaps imagine going to the moon, but how would they even start? The short answer: By breaking the problem down into parts, and then breaking the process down into steps. And then by having versions (to borrow from the analogy of software). It wasn’t Apollo 1 that went to the moon, but Apollo 11.
Such “big hairy audacious goals” (aka BHAGs) are daunting, seemingly impossible aspirations. The key here is the ability to think long-term, to project and to plan forward — much as one does when designing any other engineering-based roadmap. So, the Apollo team, like most engineers, broke things down into more doable steps of engineering; put together, those steps created the stuff of dreams.
The challenge in biology lies in breaking down the problem into steps and often reinventing the process itself. But once the desire to consistently improve performance (what Grove was suggesting in the first place) moves biology from bespoke, artisanal approaches to designed, scalable processes, even seemingly modest performance increases can make a difference. A 1 percent increase performed weekly, for instance, would lead to almost doubling in a year and a tripling in two years, and in biology, such improvements could have huge impact.
* * . *
Andy Grove did not found Intel, but he was there early and was deeply influenced not just by manufacturing methods for management but by “the law” (that the number of transistors in a circuit doubles every couple years) proposed by Intel co-founder Gordon Moore. However, Moore’s isn’t a law of physics but of economics and arose from an engineering push that continued across different technologies, different teams, different decades. It’s a law of will, imposed by man, not nature.
In biology, we’ve already surpassed Moore’s Law; the cost of genomics has come down over a million-fold in two decades. Why can’t we carry this process to other areas in bio as well? The question now isn’t whether this is possible in biology or not, as the Grove fallacy argued, but how to do it, given where we are in engineering biology today.
This article originally appeared in Scientific American.