Asimov Press

Why Nothing Can Grow on Mars*

(* = probably.) Two years ago, a friend took me out to coffee in Boston and said, “I think we can engineer an organism to terraform Mars.” Terraforming—transforming a planet so it can support life—has been the perennial dream of science-fiction authors and futurists, who, from Isaac Asimov to Kim Stanley Robinson, envisioned humanity inhabiting a multi-planetary civilization. Mars is the obvious candidate for this ambition because the other moons and planets in our solar system are even less hospitable…

Read more

2 years ago · 10 likes · Asimov Press and Devon Stork

Two years ago, a friend took me out to coffee and said, “I think we can engineer an organism that could terraform Mars.” 

Terraforming - transforming a planet so it can support life - has long been the dream of science-fiction authors and futurists who want humans to become a multi-planetary society. Mars is the obvious target for this,1 and they imagine building giant space mirrors,2 importing greenhouse gasses and dropping asteroids to make a planet habitable.3 But with modern synthetic biology, maybe it’s as simple as engineering the right organism and seeding it in the right place.

So what would it take to design an organism that could terraform Mars? Well, to start, that organism would need to survive and grow on the Martian surface. Here on Earth, life already thrives among radiation, toxins, and cold in extreme environments similar to what you’d find on the Red Planet.4 Even so, because these conditions don’t converge on Earth, few of these microbes are capable of thriving in multiple extreme conditions at once. However, biotechnology could bring all of those disparate adaptations together and make a microbe that could thrive on the barren, dry, toxic and irradiated surface of Mars. The attempt will help us understand the true limits of life, unconstrained by the extreme conditions of Earth. 

With that in mind, we’ve launched a new research nonprofit called Pioneer Labs dedicated to this question. Our mission for the next few years is to try to make an organism that could grow outside on Mars. 5

I don’t expect to succeed. For one, the surface of Mars is drier than almost any environment on Earth because of the low pressure, freezing temperatures and high salt. Anything growing on Mars needs adaptations never before seen on Earth. But even if we fail, we’ll have probed how extreme biology can get and we’ll know how much terraforming will need to be done manually before life can take over.

If we do succeed, then our microbes could start changing Mars just as terrestrial microbes changed ancient Earth. They can create greenhouse gasses to warm the planet, crack open nitrates in the soil to thicken the atmosphere, even release oxygen for living things to breathe. Progress will be slow, but turning the red planet green is the only way humans will ever walk freely on the Martian surface.

Everything that’s not a problem

Broadly speaking, there are five major obstacles that life would need to overcome to survive on Mars. Luckily, there are lifeforms on Earth that can thrive despite each of these challenges, though there’s nothing that can survive all five combined. By taking inspiration from each of these extreme organisms, we can get a sense of how to make life that can prevail on Mars.

Radiation

Radiation is usually the first consideration of space hazards,6 and it’s certainly a doozy for humans. 5 sieverts of radiation will kill half of the people exposed to it,7 and a trip to Mars and back delivers about one sievert of ionizing radiation in the form of big and fast particles that can penetrate even thick shielding.8  But microbes can survive about a thousand times more ionizing radiation than people can,9 and grow quickly enough to outpace any damage done. As a general rule, the ionizing radiation that NASA agonizes over when it comes to space missions can be ignored when thinking about microbes. 

What’s not negligible is ultraviolet light. You can protect astronauts from it with some glass or a thin sheet of metal, but if you put your microbes outside on Mars the UV will kill almost everything in mere minutes as the radiation shreds DNA and destabilizes proteins.10 That’s a problem for any prospective Mars microbe, but we can take inspiration from UV-resistant Earth microbes to see if there are ways around this. 

Deinococcus radiodurans was isolated in the 50’s by a scientist who was trying to sterilize canned meat with gamma radiation, and it’s the best studied radiation-resistant microbe.11 While it can only survive a few minutes of Mars-levels of UV radiation, there are organisms that are even more resistant to UV, like metal-loving Hymenobacter species from the deserts of Chile12 and S. solfataricus from an acidic volcano in Spain which seems able to survive Mars-levels of UV comfortably.13

This is surprising. There are no environments on Earth that even approach the amount of UV that regularly hits the surface of Mars, so there’s little reason for that level of radiation resistance to have evolved here. Still, these radiation-resistant microbes have dozens of non-overlapping ways to resist ultraviolet radiation, ranging from simple solutions like pigments that absorb radiation and antioxidants that protect against damage to having many genome copies so that they can be used as repair templates for each other.14

The best Earth UV-resistant organisms couldn’t thrive in Martian UV, but they’re part way there. I do think it’s likely that combining these adaptations with each other properly would result in a lifeform that could survive Mars’s levels of ultraviolet radiation without a problem - but that’s a hypothesis I look forward to testing.   

Toxins

Whenever you talk about going to Mars, people mention perchlorates, toxic chemicals that are used in herbicides, rocket fuel and bleach. On Earth, perchlorates are considered industrial waste, and are toxic to humans above 2 parts per billion.15 Martian soil is composed of around 1% perchlorate,16 and that’s enough to prevent plant growth.17 Additionally, any liquid water on Mars is expected to be saturated salty brine with perchlorate concentrations ranging from 15-50%.18 

But microbes figured out workarounds for perchlorates a long time ago, and Debaryomyces hansenii, an unassuming species of yeast commonly found in cheese, can not just survive, but grow in 30% perchlorate.19 It uses a variety of adaptations to get there, providing a potential roadmap to tolerate extreme levels of perchlorate.20 Other organisms are able to use perchlorate in their metabolism, turning perchlorate to water and chloride ions and getting energy out of the deal.21 NASA is collaborating with researchers at Berkeley to take advantage of this to detoxify Martian regolith so crops can grow in it.22

Temperature

Mars is cold, and anything growing there would need to be comfortable being chilly. The average temperature is about -50°C, or -60°F. But that’s about as accurate as saying the average temperature of Earth is 15°C, or 60°F. At the equator Mars can get up to 25°C, but the poles get as cold as -150°C.23 That’s cold enough for carbon dioxide to condense into dry ice right out of the air.

If we want a lifeform that can spread across Mars, it needs to be able to grow below the freezing temperature of water. The record-holder for low-temperature growth on Earth belongs to Planococcus halocryophilus, an arctic permafrost bacteria that can grow in the lab at -15°C.24 The runner-up is Psychromonas ingrahamii, a cyanobacterial algae that exists in glacial brine pockets and can grow at -12°C.25 That temperature tolerance means a significantly larger chunk of Mars is in reach, including about the equatorial third of the planet.26 

Atmosphere

The Martian atmosphere is about 1% as dense as Earth’s. It completely lacks oxygen, and instead is 95% carbon dioxide and 3% nitrogen.27 This gas mix would kill most animals, but isn’t a big deal for microbes. In fact, many microbes are anaerobic, and either don’t need oxygen or are actively poisoned by it.28 Furthermore, Cyanobacteria grow well in a simulated Martian atmosphere29 and the high levels of carbon dioxide improve the rate of photosynthesis.30

The low pressure is a separate issue, but itself is not a problem. Bacteria can grow just fine at low pressures, so long as they have the gases they need - oxygen for aerobic species, carbon dioxide for photosynthetic ones and nitrogen for nitrogen-fixers.31 Furthermore, a diverse set of bacteria from permafrost soil can grow well under Martian atmospheric pressures.32 

Water

Water is essential for life as we know it.33 There’s an organism from the siberian permafrost that’s active with the tiny amounts of water that gathers on salt crystals, and many dozens of other ways for lifeforms to conserve water,34 but that water needs to come from somewhere in the first place and it must be a liquid. No organism can create water from nothing. Here’s a map of the surface water on Mars, with white meaning no water.35

Most of the planet is totally dry, with the water frozen at the poles at -100°C36 or stuck in hydrated minerals.37 And while it was a big deal a few years ago when we had good evidence of liquid water on Mars,38 it's important to consider the state of that water - brines that are kept liquid by extremely high concentrations of salt.

This brings us to the prevailing problem with life on Mars. 

The Water on Mars is not Available

If you dumped a lake of water onto Mars, some of it would evaporate in the low pressure and condense again at the frigid poles. The salty soil would absorb the rest, binding the water tightly enough that it wouldn’t evaporate in the low pressure atmosphere. 

It’s hard to overstate how thirsty the Mars soil is. It’s salty and dry enough to suck water right out of any organism currently living on Earth. The entire planet is basically a desiccant, like those little white packets in food packages that keep them dry and prevent spoilage.39 In fact, adding salt to food is an ancient method of preservation that works by sucking up the water so it can’t be used by microbes. 

The desiccation of an environment is quantified by scientists as water activity, and it’s expressed as a number between zero and one. A water activity of 1 means pure water, while a water activity of 0 means that there is no accessible water, and any added water will be absorbed immediately. Foods with a water activity of 0.6 or lower are considered effectively immune to contamination because the lack of water prevents microbial growth.40 Dried fruits are one example, with a water activity around 0.55. This means that they will never go bad if kept properly dry. 

The known lower limit of water activity for life is 0.585, found in Aspergillus penicillioides, a fungus that lives in dust and dry paper.41 There aren’t very many organisms that can grow below a water activity of 0.7, with only 12 species known as of 2014.42 Most of these organisms grow in organic environments, like paper or honey.43 But the dry environment on Mars is caused by salts, which are harsher than carbohydrates. It follows that the salt mix and temperatures on Mars may mean you need a water activity around 0.75 or higher in order to seed life.44 

Any liquid water on Mars has a water activity below 0.5.45 There’s no way anything we know of now could grow in it. The salt would suck the liquid out of cells faster than noon in the Sahara.

NASA has looked into this. After “a deliberate and systematic search spanning several years,” they concluded in a 58-page report that “Mars is either too cold or too dry to support the propagation of terrestrial life.”46 In reaching this conclusion, NASA assumed that the limits of life are -20°C and a water activity of 0.5, both of which are very permissive given the demonstrated biophysical limits of life on Earth. 

Unlike radiation and perchlorate, where there are no analogous conditions on Earth, there are environments on our planet with water activities below 0.5, and nothing grows there. Various dried foods are one example, but another is bittern salt ponds with very high concentrations of salt.47 If nothing has figured out how to grow in those environments over billions of years of evolution, then it’s probably a pretty hard limit for normal Earth-based life that requires water to function. 

However, not all hope is lost. A theoretical Mars organism might not be able to grow in the perchlorate brines that currently exist on Mars, but it could try to create a better environment for itself in a few different ways. 

One possibility is a secreted layer that can trap heat and water, acting like a greenhouse to protect a biological community from the raw Martian conditions, and potentially thawing cleaner ice for water.48 The theoretical engineered organisms could create their own habitats, protected from the harsh conditions of Mars by their own biologically produced insulation. Biofilms already do this on Earth with sugar-based materials,49 but that almost certainly wouldn’t be good enough for Mars because the sugars aren’t good enough at holding in heat and water. You’d need to use novel materials with better properties.50 

Another organism from which we can draw inspiration is lichens, which are symbiotes of fungi and algae. They can survive extreme conditions and are able to absorb water from the air. A lichen from Antarctica once spent 18 months on the outside of the ISS, exposed to vacuum and unshielded solar UV, and came back mostly alive.51 Researchers studied that species to see if it could live on Mars, and it kept photosynthesizing under Martian pressures and temperatures, but didn’t grow.52 However, lichens survive dry conditions by drying out and waiting for water to come back,53 and on Mars, the water would never come back. I’m not sure what adaptation could fix this - maybe some kind of active water uptake or storage mechanism - but these lichens are achingly close to being able to extract enough water from the surprisingly humid Martian night and use that to propagate.54 

Ultimately, the lack of bioavailable water is the largest scientific barrier to anything growing on Mars without human intervention. Unlike the radiation, toxins, temperature and low pressure, I don’t know of any pre-existing adaptations on Earth that could get around this problem, and it’s much harder to invent new 

Legal Barriers to Terraforming

Even if we were to overcome the technical challenges of making an organism that can survive on Mars, being permitted to release that life is another issue. Currently, terraforming is prohibited by NASA policy,55 and questionable under international agreement.56 

Furthermore, terraforming Mars to an Earth-like state would probably drive any life native to Mars extinct, as earth-derived organisms would invade and transform its habitat. We’ve spent several billion dollars searching unsuccessfully for life on the Red Planet,57 but it’s very hard to prove it doesn’t exist. Taking such an irreversible step is a decision that needs to be made by nations, not by individuals.

There are loopholes that could allow terraforming to proceed regardless, but I would much rather that terraforming be unambiguously allowed by some body with the proper jurisdiction to grant such permission. Getting that agreement is just as big of a barrier to terraforming as everything technical described above, but proceeding unilaterally would be unethical and dangerous.58 Luckily, there’s a lot of cool science to be done in the meantime to answer the practical questions of how terraforming would even happen. 

The Near Future of Engineered Microbes in Space

Life already survives many of the individual extreme conditions present on Mars. But no known organism can grow in the polyextreme environment made when they’re all combined together. By mixing extremophilic traits from across Earth, we can learn more about how extreme biology can get, and what it takes to survive the irradiates, toxic, cold, low pressure, and dry wasteland that is modern Mars.59 

Our goal at Pioneer Labs is to find the true limits of life, unshackled from the environments that exist on Earth. If we make the most Mars-like microbe possible then we know how much abiotic terraforming would need to happen before it could be unleashed to finish the job. And once we know how many space mirrors or tons of greenhouse gasses are necessary, we can start making a detailed plan for really terraforming Mars. 

Even if it is impossible for anything to grow on Mars as it is now, engineered biology will be vital for supporting civilization in the stars. Polyextreme life will reduce the investment required to manufacture therapeutics, food, structural materials and chemical feedstocks vital to sustaining humans.60 This benefit will also extend to Earth, where much of the cost of biomanufacturing is wrapped up in catering to fragile organisms that require stable pH, temperature, and aeration conditions and can only grow on sterile and pure feedstocks.

In other words, Pioneer Labs is making a moonshot in humanity’s long quest to terraform Mars. Even if we fail, our home will be better because of it.

Every human has trillions of microbial cells inside of them, digesting their food, covering their skin and living in their lungs. There’s been thousands of articles over the past decade about how important the microbiome is to human health. People have been saying that treating it will change the shape of healthcare. But it hasn’t happened yet.

Today, microbiome treatments rely on fecal transplants. This is where doctors take a donor’s poop and put it into a patient, either in pill or through a tube up the colon. While fecal transplants are only FDA-approved to treat Clostridium difficile infections, doctors are investigating using them for everything from Alzheimer's to diabetes to cancer.1 But the current technique of taking poop from one person and putting it into another person is shitty because it isn’t standardized. Microbes can vary wildly donor to donor and can bring new problems along, such as multi-drug resistant pathogenic bacteria or viruses.2

To explore what might be possible, I worked with the 50Years team to explore the potential of a cocktail of microbes grown in the lab and given to patients as a first-line defense against 20 different diseases. First we dove into the research on each of these diseases, estimating a causal probability based on meta-analyses of microbes addressing disease. Then we found market sizes and patient populations for these diseases, and put all of it into a handy spreadsheet here. The result: A treatment that creates a healthy gut microbiome would be a first-line treatment for markets totaling over $500B per year and impacting more than a billion people. 

That microbe cocktail will tackle an absolutely massive market for the most vulnerable people among us. And it has the potential to drop healthcare spending in the long term by addressing diseases far closer to the cause, rather than treating symptoms or by-products generated by an unhealthy microbiome. 

So how would you make a microbiome replacement cocktail?

Bioengineers would need to identify and manufacture an ensemble of microbes that’s safe, consistent from patient to patient and tuned for healthy microbiome function. Despite $1.7B in research funding3 and over $2 billion4 in startup funding over the last 10 years, this doesn’t exist yet.5 When it does, it’s going to treat a billion people. Because the ensemble represents a healthy baseline, a doctor won’t even necessarily need a full diagnosis. If a problem is vaguely related to inflammation, metabolism or neurology? Just give them a few pills full of microbes and see if it helps.6

Let’s dive into the details and justifications for all the diseases that microbe ensembles would be able to treat.

Just like with C. diff, replacing your gut biome helps treat other infectious diseases, in the gut or not. This includes cholera, salmonella, recurrent UTIs and multidrug resistant infections.7 Microbial ensembles also prevent various bacteria from causing gastric cancer, and generally block colon cancer progression.8 

The gut is also the largest immune organ in the body, and effective communication between the immune system and the gut is critical for health.9 Raising mice without a microbiome cripples their immune system,10 and dysbiosis (an unhealthy microbiome) is just as bad.11 If your cells in your gut attack you, that can lead to an autoimmune disorder which can be treated by a fecal transplant.12 This insight reaches multiple indications, including arthritis,13 chronic fatigue,14 allergies,15 baldness,16 and aging17 on top of the obvious ones like colitis and Crohn's.18

In addition to directly treating these diseases, chemotherapies (a $10B market) and immunotherapies (an $110B market) can both be modified by the gut microbiome, and microbial ensembles can turn non-responders into responders.19 This would give a microbial ensemble a multiplicative effect on existing therapeutics.

Your microbiome processes all of the food that you eat, and how it does that is critical to regulating hunger and nutrient absorption. If your microbiome keeps you hungry, you’re going to gain weight.20 On top of that, a microbial ensemble can change how your body processes sugars, helping deal with diabetes and other metabolic disorders,21 as well as reducing the risk of the biggest killer in the world - heart disease.22 

In neurological health, the root cause of multiple sclerosis is potentially Clostridium perfringens,23 and fecal transplants have treated MS effectively in small studies.24 There hasn’t been as much study for Parkinson’s25 and Alzheimer’s,26 but there’s good evidence that an unhealthy microbiome can contribute to both. It’s possible microbial ensembles will slow or even stop these neurodegenerative disorders.27 It likely won’t reverse damage already done, but even slowing these diseases would impact tens of millions of people globally.

The immune system forms early on in life, so microbial ensembles for children would have benefits for many of the indications listed here. In addition, setting the microbiome early could prevent autoimmune disorders that cause everything from diabetes to acne. Persephone Biosciences is doing just this by sequencing the gut microbiomes of healthy infants to discover a set of microbes that could foster a healthy microbiome. While industrialization, changing diets, and antibiotics are all partially to blame for microbial changes, C-sections and lack of breast-feeding also both play a role. Persephone will initially treat babies born via C-section (27 million28 each year!), and then expand to treat all children, reversing thousands of years of microbial evolution, and preventing many of the diseases mentioned. This will be a market worth at least tens of billions of dollars annually as the science matures, given 130 million children are born each year.

Unfortunately, making an effective standardized fecal transplant is hard. Several microbiome companies have failed, including Seres, Evelo, and Finch,29 highlighting the challenges in making a good microbiome therapy. This is a new clinical space without significant precedent, and there are numerous question marks from strain selection to manufacturing and delivery. 

But for every company that’s failed, two have sprung up in their place.30 Some have even raised hundreds of millions: Seres Therapeutics has raised $537M, and Vedanta Biosciences has raised $433M. Both academic and industrial research into the microbiome is accelerating. This is spurred on by hurricane-strength tech tailwinds. Sequencing costs have dropped 1,000,000,000x in the last 25 years alongside other -omics. At the same time, AI model parameter counts have improved at least 10,000,000x. The microbiome is a high-dimensional stew of microbes, nutrients and proteins, and we need every tool we can get to understand it well enough to rebuild it from the ground up. 

While it’s been common to investigate microbiomes via in-vivo testing, in-vitro testing is rapidly gaining popularity. Gut-on-chip technology is in the lead, able to capture important interactions in vitro.31 While currently rare, this market is expected to 5x in the next 5 years.32 Other high-throughput microfluidic approaches are also contributing, allowing combinatorial testing of strain combinations and conditions to test thousands of interactions at once.33 

The development of metagenomics gave us a way to see what microbes are present in the microbiome, but it’s not enough information. Metabolomics is the next tool, as it reveals the metabolism and chemistry in our internal bioreactors. That’s led to significant advances like understanding the role of butyrate, but there are far more metabolites to map.34  There’s also been some incredible advances in metaproteomics,35 as well as direct genetic manipulation of the microbiome that provide pieces of the puzzle.36

All of these new techniques generate an enormous amount of high-dimensional data, which is hard to process and harder to generate appropriate models from. But artificial intelligence is well-suited to help us build effective synthetic microbial communities that can replace unbalanced microbiomes.37 And AI technology is getting vastly cheaper.

The first company to properly take advantage of these technologies and make an effective and scalable synthetic fecal transplant will have a massive first-mover advantage. Doctors will try it to treat everything named above - and more. Positive results will get that same microbial ensemble approved for more indications beyond the first, and negative results will provide valuable data to improve the treatment.  There will likely also be challenges in manufacturing initially, and so the first mover will have massive demand that facilitates scale-up financing. These factors mean that followers will have to spend far more capital per indication, both for running studies and for manufacturing microbial ensembles. Being the first mover could be exceptionally profitable. 

I learned to quit by giving up on my first amazing idea. It was a very valuable lesson on not wasting time.

In the second year of my PhD program, I was working on making targeted antimicrobial peptides to kill specific pathogens. Five Harvard professors had given their stamp of approval during my qualifying exams, and I was ready to do science that would change the world. Or at least create some new drugs that would probably fail in phase two of clinical trials,1 but that’s the life of a biotech scientist. But there’s a huge gap between “technically possible” and “I can do this.” At that point I made one of the best decisions I’ve ever made. 

I gave up. 

I apprenticed myself to a senior postdoc and learned how to do genetic code expansion instead.2 That project gave me a plethora of skills and good ideas, one of which quickly developed into my PhD.3 By killing my first idea, I had a five-year PhD instead of the eight years-or-never it might have been if I’d tried to hold onto my first project. 

Killing projects frequently is really important, because it lets you try new things that might work better.

We all aspire to persevere in the face of adversity, and we might have promised somebody results, or be convinced that this project is vital to our career. But in my experience you can always get around those barriers one way or another. The real reason people don’t quit is that they focus too hard on the challenge in front of them, even when their perseverance would be better off pointed somewhere else. Nobody wants to admit they made a mistake, and giving up feels like admitting that you’ve wasted time and effort,4 even when it’s a reasonable response to new information. 

People who build things put their self-worth into what they build, and giving up on your project is like giving up a part of yourself. Being proud of your work is incredible, and a core part of my identity is my science and what I build with it. But it’s a trap to get attached to any specific thing you’re working on. When your self-worth is tied up in a specific project that’s going badly you’ll twist the facts to preserve your own ego and persevere long beyond the point of reason.5 You will get increasingly upset when it’s not working, and that’s a quick path to depression or burnout. 

The solution is to learn the right time to give up. That will usually feel like it’s too early, but it’s also not at the first sign of trouble. The thought that got me to quit my first idea was “I never would have started if I’d known where it would go.” Now whenever I hit a block I ask myself if troubleshooting will take as much effort as starting something new. If so, it’s usually time to kill it and try to find something better. Very few things work on the first try, but the good ones will typically get moving with only simple troubleshooting.

My advisor, Professor George Church, framed his research as looking for problems that seemed impossible but weren’t really that challenging. Some really significant projects are incredibly difficult, but others are very straightforward. The best way to find the places where rapid progress is possible is to try several ideas and move on quickly when they don’t work out. 

My PhD project was about expanding the genetic code of Bacillus subtilis, and it didn’t work immediately. But it did only take about two weeks of straightforward troubleshooting to figure out that the key was seeding exponential phase cells into minimal media and then growing them overnight.6 That discovery was what gave me traction7 on that project.

Traction is the place where you have things you want to do and nothing is stopping you, it’s awesome. Everything you do spawns new directions. The experiment that gave me traction is ⅙ of figure 1D, but it allowed me to build the rest of the publication in about a year and a half of joyful experimentation. Don’t quit a project when you still have traction.  

But, eventually you’ll lose traction, often by reaching the limit of your tools or running out of questions to ask. Quitting after you’ve found traction can be harder than quitting before you ever get there. Towards the end of my project I didn’t want to accept that I’d gone as far as the project could take me. Then COVID lockdowns happened. Getting booted out of the lab for six months at that time was pretty lucky, since it forced me to write everything up and think strongly about what would come next. That accelerated the rest of my life by about a year. 

After graduation I joined a friend’s biotech startup, where I had a somewhat rude awakening. A lot of the time, you can’t abandon projects. In industry you will always have people like CEOs and investors and commercial partners who expect results on specific projects. This can also happen in academia with grants and collaborators, but I had to go to industry to learn how to deal with projects you simply can’t quit. 

The solution is that there’s always more than one way to get something done. You might be married to the goal, but you’re not married to any specific attempt to get there. Figure out a new approach and try that instead. Trying new things lets you see the bigger picture and sometimes the solution will end up being a combination of your first and third attempts. 

Ultimately, changing what you’re doing is emotionally hard. Seriously considering a new idea is scary, because it means abandoning something comfortable for something new. But to me, one of the most valuable lessons of graduate school was learning to embrace the potential of something new and accept the risks that came with it. 

That lesson is valuable across life, and for reasons beyond something not working. I’m leaving my current (excellent) job engineering gut probiotics to move across the country and start a nonprofit focused on building biotechnology for space. I have an opportunity to engineer organisms that can sustain life in the stars, and that’s too exciting an opportunity to pass up. 

In the end, while not all change is improvement, all improvement is change. 

I’d like to thank Dan Voicu, Maggie Chen, Niko McCarty, Xander Balwit, Gabriel Filsinger, Sarah Scheffler and Anjali Kayal for providing feedback on this article, as well as the Ideas Matter Fellowship, without which it wouldn’t have been written. 

But one of my touchstones to understanding life is that “Nothing in Biology Makes Sense except in the Light of Evolution”1. To comprehend the purpose of anything biological, you need to understand what pressures guided its evolution. Over the last few billion years this process has shaped every living thing - and the planet Earth.

A key insight is that organisms are not passive participants in evolution. There are a plethora of adaptations and traits that actively guide evolution, controlling how adaptive diversity is generated, spread and selected upon. Some simple examples are mutation rate, genome size and sexual reproduction. But if we dive deeper, we see that the fundamental natures of deep biological concepts - like gene regulation and genetic architecture - are optimized for evolvability.

These are meta-adaptions, traits that change how further traits are acquired, and understanding them helps us understand why life functions the way it does. If we can figure out the purpose behind evolvability adaptations, we can start mindfully engineering them to not just control an organism, but control how it evolves. This kind of knowledge will be vital as bioengineering moves towards whole-organism design.

Furthermore, being able to control evolvability unlocks new tools. Supercharging evolution would let us quickly adapt life to an incredible array of conditions, like the insides of bioreactors, polluted mines and even other planets in our solar system2.

Just as important would be ‘freezing’ a designed species to prevent evolution. We don’t want our bioproduction or carbon sequestration strains developing beyond what we’ve built them to do3. If we can one day grow a house, we don’t want that house evolving away from its blueprint - because it almost certainly won’t evolve to become a better house4.

If we as a species want to get good at engineering the self-replicating nanotechnology of life, we need to master the force that shapes its development - evolution.

Evolvability is Diversity Acquisition

Evolution happens by selection on existing genetic diversity. Therefore, if you want to evolve faster, you need to generate diversity faster so that the environment can select upon it. This means you want to generate a lot of beneficial and functional heterogeneity very quickly and then hold onto it, while trying to avoid deleterious mutations.

Traits that increase evolvability are co-associated with advantageous diversity that they helped generate, so they get selected for5. Over time more evolvable organisms adapt faster to changing circumstances and fill newly available niches before any competitors. This process is independent of any specific environment, so it’s universal. Traits that encourage faster evolution are selected for.

The basic version of evolvability is enhancing how quickly you accrue diversity by tweaking mutation rates and encouraging recombination. Every organism precisely tunes mutation rate to hit the sweet spot where you’re gaining the right amount of diversity but not breaking too many things6. Recombination happens via sexual reproduction and horizontal gene transfer, but it’s limited by the existing pool of diversity so there’s an upper cap to how much diversity you can gain. Those are the easy evolvability adaptions.

The advanced version of evolvability is shaping the underlying systems so that random perturbations are more likely to be adaptive and less likely to be deleterious. This has deep implications for the basic way that biological systems work.

Reliability vs. Robustness

In engineering, we think a lot about response to perturbation. You design things to not diverge from specs even in noisy conditions. This is reliability - you want your system to keep functioning the same way even as individual pieces degrade7. But eventually you’ll hit a failure point and the entire system will fail dramatically.

Reliability is not a good feature for an evolvable system. An evolving organism should change fluidly as individual components are swapped out or added, with variable but still functional outputs coming from small genetic changes8. Evolvability wants genetic diversity to translate well into phenotypic diversity, without significantly disrupting the overall system.

That last part is important - an evolvable system needs to be robust. If some parts stop doing their original job, other parts will pick up the slack to keep the organism alive. This robustness increases the space of mutations that are available, opening up the greatest diversity of genotypes and phenotypes. Not only does this mean you’ll get more good mutations, but you’ll also be more capable of holding onto a pool of neutral or weakly negative mutations that might become advantageous if the environment changes.

These two design principles seem opposed. How can you have a system that is both translatable and robust? You want your organism to survive the maximum possible range of mutations, but you want all of those mutations to have real, phenotypic effects.

Signaling Networks

To me, the best examples of robust and translatable systems are eukaryotic signaling networks. These do a huge amount of data processing in higher organisms, controlling gene expression - and thus cellular behavior - in every cell in our bodies.

Here’s a graph representing signal transduction in the yeast Ras-cAMP signaling network9.

This is somewhat small as these kinds of networks go10, and it looks a lot like a recurrent neural network. Indeed, modelling it as a neural net is a recent and successful approach to understanding how these systems compute biological signals11.

Beyond wasting hundreds of PhD-years of time and fueling the reproducibility crisis, the complexity and feedback in these networks means that very few nodes are truly essential for any given response. However, a change in any node will likely cause some shift in every output.

This system is a nightmare to try to understand, but fabulous for evolvability - any given change will have a phenotypic effect across multiple outputs, but is unlikely to completely break anything12. Combine that with sexual recombination to mix-and-match different mutations and you’ll maximize the adaptive potential of what diversity you have.

Evolvability is not the only reason that mammalian signal processing is complex, since it also the system harder for parasites to hijack13 and probably other things too. However, I think that a primary reason for this complexity is improved evolvability.

Furthermore, biology won’t always converge on something like a recurrent neural net as the best way to do signal processing if it doesn’t need the extra evolvability. Bacteria primarily use pretty simple two-component systems for regulation, and they have some fantastically complex behavior that arises from much simpler one-to-one networks14.

The differences arise because bacteria and mammals exist in very different evolvability niches. They have very different strategies for acquiring and storing diversity.

Genetic Architecture

To dive into this idea further, let’s compare the evolvability strategies of E. coli and Homo sapiens, then talk about how that determines what their genomes look like.

E. coli live in huge populations, divide quickly and acquire diversity in response to a changing environment. They do this by relying on spontaneous mutations and by picking up free-floating environmental DNA. This strategy has a low hit rate, but across large populations you’ll usually end up with something adaptive15.

On the other hand, Homo sapiens live in small populations, divide slowly and store diversity for a long period of time. We don’t reproduce fast enough to generate selection in response to an environmental pressure, instead banking rare occurrences of diversity so there’s an existing pool to be selected upon when necessary.

These differing evolutionary strategies shape an incredible number of the basic traits of these organisms. E. coli need to be able to divide very quickly when times are good, and therefore reproduce clonally and keep their genome size small. Humans reproduce so slowly that genome size isn’t particularly limiting16, but we store a lot of diversity and mix and match it through sexual reproduction. Even though we only have five times as many genes as E. coli, our genome is a thousand times larger.

Much of that extra DNA isn’t functional. Almost all of the human genome is a messy “junk” of transposons, broken paralogs and regulatory elements, recombining and drifting without being essential or even especially important. Most of this DNA gets transcribed anyway17, even if it doesn’t do anything18. It just sits there for tens of thousands of years, accumulating and storing diversity. Every now and then an ‘orphan gene’ pops out of this morass of DNA, with no homologs in other species19. The same thing happens with regulatory elements, often directly caused by freely hopping transposons20.

While orphan genes are cool, they are rare. But a few billion bases not under selection is a lot of space for random things to happen. More common is that a gene or regulatory element will duplicate and remain mostly intact, drifting far enough to create another member of the same family while never fully becoming nonfunctional. Then it can fulfill a similar function as the original protein in a different context, and the process is more likely to happen again since there are more versions to duplicate. The incredible diversity of GPCRs are a great example of this happening over evolutionary time21.

There are more examples of evolvability literally everywhere in biology, but I just want to hit a few more high points that lie in my area of expertise.

• Introns provide a large target so domains can get easily swapped between proteins. Most recombination events anywhere in several-thousand-base intron will produce a properly spliced protein with a new combination of domains. Without introns, you need single-base precision in your recombination.

• The nucleus creates physical separation of transcription and translation, allowing those steps to be differentially regulated in eukaryotes. This is one more lever that can be used to generate diversity - if it is beneficial to upregulate a specific developmental marker at a specific time to grow larger lungs or something, that can now happen in multiple ways.

• The above is also true for post-translational modifications.

• Plants have a suite of adaptations to encourage genome plasticitiy22, enabling interspecies hybridization and tolerance of variable genome ploidy. For example, invasive weeds often have incredibly plastic genomes that allow them to adapt quickly when invading a new niche. They hybridize with everything and their genomes balloon, then the bloat slowly gets purged down to something well-adapted for the local environment23.

• Bacteria package similar genes together into operons so they can be treated as a functional unit that is easy to regulate as a unit, meaning that one mutation can easily shift the activity of an entire pathway.

• Viruses are the masters of evolvability through incredibly rapid replication & rapid base-pair optimization. They often stack multiple genes atop each other in different coding frames to reduce genome size24.

I’m sure that there are similarly fundamental systems across biology that are heavily influenced by evolvability. Some examples might be how morphogen gradients work and how brain architecture is encoded. But I’m an expert in bacterial genetics and have dabbled in aging research25, so I don’t feel qualified to speculate on those topics.

It’s hard to think about the effect size of any evolvability adaptation. But most of them open up whole new dimensions of diversity, so the outcome is rather nonlinear. We’d also expect more evolvability adaptations in larger organisms that replicate more slowly. They need all of the advantages they can get to keep up in the red queen’s race.

Engineering Evolvability

I hope you will agree with me that evolvability is not just a knob to turn up or down, but a fundamental feature of biology that must be taken into account when engineering life. That insight doesn’t necessarily limit what kinds of engineering we can do. If anything, knowledge of evolvability gives us a clearer view of the implications of biological research.

For example, one of the main limitations on the speed of animal embryonic development is the time it takes for genome duplication26. If we want to engineer organisms with faster gestation times, reducing the genome size by taking out all of the nonfunctional DNA might be a good way to do that. By understanding why that DNA exists, we know what we’re sacrificing. It probably wouldn’t be very effective to do selective breeding on cattle with a minimal genome because we’ve removed a lot of their banked diversity.

Similarly, the complexity of mammalian transcriptional regulation isn’t vital for functionality, but if you engineer a simple and direct synthetic signaling network, it will be much more likely to break instead of adapt. This is probably a good thing if we’re growing houses and you’d rather have something die instead of grow out of control. In order to engineer biological systems that are reliable instead of robust, decreasing evolvability is a good thing.

On the other end of the spectrum, directed evolution is an important way to develop new biological technology27. Optimizing evolvability has revolutionized this field, yielding more powerful results by increasing the search space28.

These developments haven’t reached whole-organism engineering - but that’s where they’ll be the most impactful, especially when we start synthesizing genomes. The first big target is likely to be biomanufacturing, since none of our bioproduction organisms evolved to do well inside giant steel tanks full of industrial solvents29, fed rich nutrients to continually produce bioproducts. We’ll need to target evolution towards that end. The more elements of evolvability we include in that process the better the result will be.

This principle is true beyond bioreactors. If you want to take life to space, you’re going to run into environments and raw materials that no life on earth has evolved for, such as abundant perchlorates and unshielded ultraviolet radiation30. These are novel challenges to biology, and we’ll need to build organisms for rapid evolution to spread life to the stars.

Just as important as accelerating evolution is learning how to slam on the brakes. If you can slow evolution to near-zero, you prevent adaptation away from design parameters. We need to biocontain strains before they can be used outside of the lab31, and decreasing mutation rate, simplifying genome architecture and removing recombination mechanisms are all important steps towards that goal. Next-generation biocontainment might even rely on low evolvability as a primary way to make an engineered organism noncompetitive in the wild.

Life Understanding Life

Finally, I think that evolvability is an essential part of understanding biology. The best way to understand the purpose behind a given trait is to understand how it evolved. The best way to understand many traits together is to understand how their interactions shape their evolution and capacity to adapt.

Many synthetic biologists repeat the Feynman quote “What I cannot create, I do not understand.” Part of engineering life is not just being able to design it de novo, but to understand and engineer the actual forces that governed the creation of life as we know it. Synthetic biology as a discipline should consider evolvability of both natural and designed systems, and always be aware of the trade-off between reliability and robustness.

Subscribe now

Thanks to Merrick Smela, Anjali Kayal and Sarah Scheffler for providing feedback on this article.

30

Asimov Press

Why Nothing Can Grow on Mars*

(* = probably…

Read more

2 years ago · 33 likes · 1 comment · Asimov Press and Devon Stork

Viscosity and Fluid dynamics

Viscosity is simply a property of a fluid. Water has a very low viscosity - if you drop a rock into water it will quickly sink. Thick oil has a much higher viscosity - the rock will still sink, but more slowly.

Viscosity creates two regimes of fluid dynamics. At high viscosity, all motions are damped and fluid flow is smooth and constant. At low viscosity, motions are undamped, inertia dominates and fluid flow is chaotic and turbulent.

Which regime you are in depends on more than the viscosity of the fluid - the flow speed, scale and density of the system also matter. This leads us to Reynold’s number (Re). Reynold’s number is defined as the ratio of inertial forces to viscous forces1 and is a way to tell if movement in your system is dominated by inertia or viscosity. A clever derivation2 gives us a simple way to calculate Re:

Where ρ is the density of the fluid, u is the fluid flow speed, μ is the viscosity and L is the characteristic dimension, effectively the scale on which you’re working. The two terms that depend on the fluid are ρ & μ, and for water ρ/μ is about 0.01 cm^2/second. Therefore the Re for an object moving through water scales directly with the size of the object and how fast it is going.

This should make intuitive sense - bigger, faster objects care more about inertia and less about viscosity. Blue whales are very big3 and it makes sense they are less effected by viscous drag than swimming people are. In water the blue whale Re~4x10^8, while Re for a swimming human is about 4x10^64. This means a human needs to exert more continuous effort to keep moving, because viscous drag slows them down more proportionally than it does a whale.

As we keep going down in scale, this process continues. An extremely small fish in water may have an Re between 1-10, and it will quickly slow to a halt unless it is continually swimming.

What is the Reynold’s number of a cell?

This effect is extreme on the scale of a cell. Most bacterial cells are about a micron across and swim around at about 20 microns/second5. Thus, the Re of a typical bacterial cell in water is around 10^-4. If they stop swimming, inertia will carry them 0.1 angstroms, or 10^-5 of their size. Because of the tiny dimensions and (relatively!) slow speed of these cells, they are effectively without momentum.

It’s difficult to understand the implications of not having inertia. The example Purcell uses is to imagine a person trying to swim through a pool of molasses without being able to move any part of their body faster than 1 cm/minute. They can’t build up any inertia to break through the viscosity of the molasses, and are well and truly stuck.

This gets weirder - if you throw out inertia, time stops mattering. The only thing that matters to your current velocity is your current force, and all past forces are irrelevant. Acceleration is instantaneous, and swimming requires specialized movements.

To illustrate this, consider a swimming scallop. It jets along by slowly opening it’s shell to draw in water, then quickly closing it to expel water, building up inertia to make progress.

Figure credit6

This movement would get the scallop nowhere at low Re - the opening motion would pull it backward just as far as closing pushes it forward. We can extend this principle to say that swimming at low Re cannot be reciprocal. It must be a non-time-reversible change of configurations where the motion is different backwards and forwards. The standard swim kick just jackknifes you in the water, it doesn’t make you go anywhere. This is sometimes referred to as the “Scallop Theorem”.

How do you swim at low Re?

But, cells still swim, and do it pretty well. How? There are two primary solutions to the Scallop theorem.

Figure credit7.

First, a circular motion is non-reversible. So if you attach a flexible filament to a rotational motor you get a flagellum, a propeller-like structure that can push against liquid to move at low Re. Unfortunately, our intuition of propellers isn’t very useful here, since a flagella at low Re doesn’t create a backwards jet of fluid like a large-scale propeller does. Instead it slides itself through the liquid more like a screw into wood, though with a lot more slipping.

Second, a cilia uses a back-and-forth beating motion with a flexible filament. This is non-reversible, as the filament bends one way during the first half of a stroke and the other way during the second half of the stroke. This bending motion pushes against the liquid to provide forward force, though most of the force is sideways and reciprocal.

Both of these approaches have evolved multiple times across biology, and are used by nearly every organism from bacteria to human white blood cells. They’re not necessarily energy-efficient, with Purcell’s calculations estimating that about 1% of the energy that goes into these systems results in forward motion. But overall this is very little energy compared to the energetic cost of continually synthesizing proteins, lipids and nucleic acids.

Out-swimming diffusion

How fast do bacteria have to swim for it to be useful? What do they gain from swimming?

It makes sense that swimming would let you gather more nutrients, in that you’re covering more area and finding more nutrients. But diffusion on the scale of a single cell is fast. To increase the amount of nutrients a cell can absorb would require really high speeds - Purcell calculated out that a cell in a uniform field of nutrients would have to move 700 microns/second to increase the available nutrients 10% (see page 10 for his calculations). That’s more than ten times as fast as bacteria are capable of swimming.

But what if the field of nutrients isn’t uniform? How far do you have to go to see a reasonable change in nutrient conversation? If you can go in a straight line for a while you’ll outrun individual molecules because diffusing molecules are not fast at going long distances, because they turn around and double back a lot. To diffuse a distance l, a molecule takes time t = l^2/D, where D is the diffusion constant. Diffusion velocity over a distance l is therefore v = l/t = D/l. Notice that the equation for velocity is divided by distance - diffusion is slower over longer distances because diffusive motion includes a lot of doubling back.

The diffusion constant for glucose in water is about 600 microns^2/second8. So if a cell moves at 20 microns/second, it moves faster than diffusion at distances of 600/20 = 30 microns and over. That’s how far you have to go in a straight line to escape your local diffusion environment.

Here’s a classic diagram of an E. coli swimming from Howard Berg’s website. You’ll notice it goes about 30-50 microns on most of the straight-line paths.

Figure credit9

If you’re more interested in how bacteria do chemotaxis, you can find a few sources explaining how they do a ‘directed random walk’ to move towards higher nutrient concentrations.10

Symposium on Biomaterials, Fuels and Chemicals (SBFC) May 4-7, 2025. Half of talks (two-track conference).

Synthetic Evolution, Engineering and Design (SEED), May 30 - June 2, 2023. All talks.

Synthetic Evolution, Engineering and Design (SEED), May 2-5 2022 . All talks represented.

The great COVID gap here.

GP-Write and SC2.0 November 11-14 2019. 41/63 talks are represented due to speaker permissions. Contact me directly if you want the notes on the rest.

Synthetic Evolution, Engineering and Design (SEED), June 24-27 2019 . 32/51 talks talks are represented due to speaker permissions. Contact me directly if you want the notes on the rest.

Boston Bacterial meeting, June 6-7th 2019 All talks represented.

ACS Annual Conference, March 18-21 2018. Only 35 selected talks from Chemical Biology sections are represented. Notes are lower-quality than more recent conferences.

The Specificity Problem

A big problem in biological systems is that a lot of potential substrates look alike. A necessary step ahead of protein synthesis is attaching amino acids to transfer RNA, or tRNA. But the enzyme that does this, the tRNA synthetase, has to discriminate between twenty different alternatives, some of which are very similar. For example, look at the structures of phenylalanine vs. tyrosine or serine vs. threonine.

Since the reaction happens down near the alpha carbon where the molecules are identical, it’s hard to prevent a synthetase that binds the wrong substrate from carrying out the wrong reaction. And yet somehow, all the synthetases have an error rate (which we’ll call f) significantly lower than one in a thousand.

What does an f < 0.001 mean from a kinetics viewpoint? Let’s look at the example of serine tRNA charging using standard Michaelis–Menten kinetics. First the enzyme (E) must bind the substrate (S for serine or T for threonine) with dissociation constants of KdS and KdT respectively. Then the reaction happens at some rate m to produce either the correct product P or the incorrect product P’. Since the reactive groups of the substrates look identical, m is the same for serine and threonine. All the specificity lies in the binding constants, the enzymatic function cannot differentiate between substrates.

If we’re at steady state, with equal concentrations of S & T, the error rate f is simply the ratio of KdS & KdT . We can expand upon this by connecting the binding energy to the dissociation constant with ∆G = RT ln(Kd) to give us a formula that connects energy of binding to error rate:

If f = 1/1000, then ∆Gt−∆Gs must be a bit more than 23 kJ/mol. This is a lot of energy - especially for discriminating between substrates that are different by a single methyl group. These energy demands get even more ridiculous when you look at polymerases, some of which have error rates of one in hundreds of millions of reactions. This implies a ΔG difference of 35-45 kJ between binding different bases, which is just ridiculous.

Breaking down the pathway

Of course, we’ve simplified this system dramatically to use Michaelis–Menten kinetics. What do we need to add back in to understand the real-world specificity? Well, there’s a known intermediate step where the amino acid is conjugated with AMP (which we’ll denote S*), yielding the following reaction pathway.

This doesn’t obviously help – we’ve just expanded out the final arrow towards the product. The ratio of incorrect to correct products will be present in both the E-S state and E-S*, proceeding to E-P.

However, it’s known the phosphorylated amino acid can unbind without proceeding to the final reaction step, and there are high-activity enzymes whose sole purpose is to remove the AMP from these byproducts, recycling those unbound intermediates back to substrate. Adding this mechanism makes the pathway look like this:

The incorrect substrate T also goes through the same process, and will have a Kdt2 of its own. Since those Kd’s are still dissociation constants for the substrate, it is likely that Kdt2 will be smaller than Kds2, meaning more of T than S will exit at that second binding step. And since the pool of E-S* is already biased towards the correct product by the ratio f, allowing that pool of E-S* to undergo another unbinding step biases the pool of substrate as if we were doing the initial binding step again. To be more precise, if Kdt2/Kds2=Kdt/Kds, then the above reaction pathway has specificity f squared. If the ratios aren’t equal, then we still get the sum of the ∆G’s of binding.

By adding an extra step to the enzymatic function, we have squared the specificity or halved the energy requirement for a given specificity.

However, there are a couple of very important details before we can present a generalized version of kinetic proofreading.

1. Any S* or T* binding to the enzyme bypasses proofreading. Therefore, the conversion of S* or T* back to S and T must be extremely rapid. This ensures that the only way to make E-S* is by E binding to S.

2. The conversion step E-S -> E-S* needs to be irreversible or at least heavily forced. This separates the two proofreading steps, guaranteeing that the substrate bias gained by the first binding step can be improved upon by the second binding step. Otherwise, the increased proportion of E-S* would revert to E-S compared to a much lower proportion of E-T* reverting to E-T.  The only way to get irreversible steps is to spend energy, so the conversion step needs to be an energy-expending step.

From here, we can state a generalized version of kinetic proofreading for substrate A of enzyme E to make product Z by way of intermediates B, C….

Each additional proofreading step will add another layer of proofreading, and thus another power to the exponent of specificity. With two intermediates and three levels of proofreading we would cube f, the specificity.

Implications and Examples

Now that we have a general description, let’s take a step back and think about this for a second. You can square your specificity by spending energy to add a step to your reaction, but only if you allow the modified substrate to potentially unbind afterwards, thereby wasting that energy. These reactions are beating the kinetic and thermodynamic limits on substrate specificity by spending energy on the problem to force an extra proofreading step. They’re improving faithful information transfer and reducing entropy by spending extra energy on the problem. It makes a beautiful kind of sense that the energy to make the conversion step unidirectional is required in order to increase the specificity of the enzyme, and that the energy is often lost by the dissociation of both correct and incorrect substrates.

Finally, I want to talk about how general this principle is. I’ve already discussed it for tRNA synthetases, which are responsible for the first step of accurate protein synthesis. It also occurs in the next step of protein synthesis, where charged tRNAs bind to mRNAs in the ribosome to generate a new protein. The key to this is a universally conserved protein called EF-Tu, which binds tRNAs and is required for tRNAs to bind to the ribosome. But EF-Tu must cleave a GTP and dissociate before the peptide chain is extended, and the tRNA can unbind after EF-Tu departure but before the peptide is extended. The overall schema looks familiar:

The more you look, the more commonly you see kinetic proofreading in cases where accuracy matters. It’s no surprise to see DNA replication on the list, though there are some other polymerase-specific ways that polymerases have increased specificity. The first energy-expending step attaches the new nucleotide (here dAMP) to the growing strand. However, it can then be cleaved off and dissociate in the proofreading step, and the procession of the polymerase to the next nucleotide is the final step.

Kinetic proofreading is also an important principle in eukaryotic signaling cascades, especially in the immune system to determine true-positive antigen binding and avoid off-target immune function. Both DNA damage repair mechanisms and homology-directed recombination use kinetic proofreading to ensure they’re attaching the correct pieces of DNA. The principle of putting in an exit point after an energy-expending step to increase specificity is present in myriad and various biological systems.

This work is a prime example of the last century of biology, where advances were often made by sitting down and thinking up a single beautiful experiment. J.J. Hopfield’s paper, cited above, lays out kinetic proofreading in a longer and more detailed fashion than this post and doesn’t contain a single experiment. However, it prompted several experimental papers in subsequent years that simply and beautifully demonstrated kinetic proofreading in real systems.