C&EN’s 2020 10 Start-Ups to Watch

Credit: Aryballe

Sam Guilaumé, CEO of Aryballe

Scientists have long sought to replicate the remarkable ability of the human nose with dispassionate equipment. But as any master perfumer can tell you, such efforts have a long way to go.

The French start-up Aryballe reckons it is at least narrowing the performance gap with its electronic nose, which combines biochemical sensors, advanced optics, and machine learning to collect and identify thousands of odors.

Nonhuman identification of smells today is largely the province of systems based on gas chromatography and mass spectrometry. Although accurate, they feature bulky equipment, and the associated data analysis can be laborious and time consuming. Like panels of human tasters and smellers, they can also be expensive.

“Aryballe’s technology retains the accuracy of such big machines but can be used in a handheld device and potentially something the size of a watch and is much cheaper,” CEO Sam Guilaumé says.

Aryballe’s sensor, about the size of a paper clip, features peptides grafted onto a 50 mm² silicon chip. Mimicking the olfactory sensors of the human nose, the peptides react with volatile organic compounds emitted from, say, a sample of coffee or a vial of perfume. During the reaction, which is transient, the peptides absorb some light. These changes in the light absorbance of the peptides are measured by an array of optical sensors. The optical sensor readings of the changes in light on the surfaces of the peptides are then uploaded to Aryballe’s server, where they are matched with the patterns of known smells.

Using light to sense smell

Aryballe has created a sensor about the size of a paper clip that can identify thousands of smells. The sensor comprises different peptides grafted onto a 10 x 5 mm silicon chip. The peptides react variously to different volatile organic compounds—the chemicals that cause smell. Different smells cause a difference in light absorbance across the surfaces of the peptides, creating a signature pattern for each smell. This smell pattern is captured by optical sensors and then uploaded to be matched to the pattern of a known smell in Aryballe’s database.

Credit: Yang H. Ku/C&EN/Shutterstock

The idea for Aryballe emerged when Guilaumé, an electronics entrepreneur, started chatting about sensors with his neighbor Tristan Rousselle, a pharmaceutical scientist and entrepreneur. “We immediately had a great connection,” Guilaumé recalls. Rousselle had been using organic compounds to identify the presence of viruses in a liquid. He wondered whether a version of this approach could be applied to smell, and so Aryballe was born, founded by Rousselle.

The firm has raised about $20 million from investors as diverse as International Flavors & Fragrances (IFF), which is a New York City–based developer of flavors and fragrances, and the carmaker Hyundai Motor.

Initial applications for a customer like IFF will be to identify fragrances or provide, for example, a quick pass or fail to incoming raw materials. For food and beverage firms, Aryballe’s technology can characterize and measure odors like the smell of coffee.

For the auto industry, Aryballe’s sensor can identify smells that indicate problems with a car, such as the burning of a plastic coating on an electric cable. The technology is good enough to determine whether a burning cable is in the entertainment console or something more critical, like the braking system, Guilaumé says.

The firm already has enough cash in place to take its next step: fit out a 200 m² production suite in Grenoble, France. Aryballe aims to ramp up production of its sensors in the second half of 2021. Its goal is to ship 2,000 sensors per year by the end of 2022. The firm also hopes to introduce a prototype sensor capable of fitting into consumer devices like smart watches next year.

Aryballe has set up offices in New Jersey and Germany and is about to open another one in Seoul, South Korea, to get closer to existing partners and potential ones in Asia. The next few years promise to be busy ones for Aryballe’s staffers as they fine-tune their technology and roll it out commercially.

Guilaumé says employees are embracing the fast pace. “We like science. We are like kids in a candy store,” he says.

Credit: Culture Biosciences

Will Patrick (left) and Matt Ball, founders of Culture Biosciences

Will Patrick and Matt Ball are friends who share a passion for the potential of biomanufacturing to improve the environment and human health. But they’re not biologists; they’re engineers who make hardware and software. More than that, they’re engineers who yearn to solve problems.

Patrick had learned to combine biology and engineering through design programs at the MIT Media Lab and the software firm Autodesk. Then, inspired by Patrick’s mom—a counseling psychologist—the two friends drove around San Francisco to meet with people at synthetic-biology companies and talk with them about their problems.

What they learned is that companies know how to craft an organism that can produce a useful molecule; their main sticking point is fine-tuning the organism and developing a manufacturing process that allows the organism to thrive and the firms to make money.

“We saw a business opportunity to make it convenient and simple” to bridge the manufacturing gap, Patrick says. In 2016, Patrick and Ball founded Culture Biosciences to speed companies’ time to commercialization.

These days, researchers at firms small and large want to engineer bacteria, fungi, yeast, or mammalian cells to make molecules for drugs, materials, food, and crop protection.

Optimizing biomanufacturing

Culture Biosciences provides cloud-connected bioreactor capacity to enable biomanufacturing firms to run hundreds or thousands of optimization experiments. Scientists at a customer firm send to Culture several modified microbe or cell strains plus information on the scaled-up manufacturing system. The customer can then go online to see data and analysis from sensors on Culture's bioreactors and from sample tests and request changes to the reactor conditions. At the end, they receive final data and the physical product from the finished process.

Credit: Yang H. Ku/C&EN/Shutterstock

But getting an organism or cell ready for industrial manufacturing requires a well-stocked lab with many fermentation bioreactors, people to run them, and the expertise to manage and analyze buckets of data. Creating that testing infrastructure is difficult, time consuming, and expensive.

“That’s not where the attention should be,” Patrick says. “Instead, companies should be working on what products will generate a huge impact on the world and generate revenue.”

Early on, a company scientist needs to create only a few versions of the organism to answer the question “Can it make this substance?” If the answer is yes, the next stage is trying out hundreds or thousands of variants to obtain an organism that is genetically stable, will produce high yields, and can perk along happily in an industrial-scale fermentation tank. It’s those latter steps of microbe and bioreactor optimization where Culture Biosciences helps.

Customers ship Culture small amounts of their microbes. They also provide details about their proposed fermentation environment, such as the growth media, operating temperatures, and pattern for pumping in sugar and oxygen. Culture then replicates those conditions in its test-scale bioreactors.

A typical customer might rent space for 20 bioreactor runs per month for 12 months, Patrick says. A firm working on a drug ingredient might even use hundreds of reactors at a time in its quest to be first to market.

Unlike standard lab-scale bioreactors, Culture’s don’t have buttons or dedicated computers; instead, they are connected to a facility-wide computer system that runs them all.

Control systems, sensors, and cameras send real-time information from each reactor to a data repository that lives in the cloud. Culture’s software analyzes data from each reactor and the images of what’s going on inside.

Customers log on to the cloud to check how their microbes are doing. Are they metabolizing media too fast or raising the temperature in the reactor? Are they starved for oxygen? Is the broth getting foamy or weird looking? If so, Culture can adjust settings on the bioreactors at customers’ request; the firm is building online tools to allow customers to make their own changes in real time.

Meanwhile, Culture’s scientists take samples throughout the experiment to find out how many cells are present and how much of the desired product they’re spitting out. Culture can send samples back to customers by freezing them and shipping them overnight.

Culture is currently nurturing modified microbes from a number of fast-growing synthetic-biology companies, including Zymergen, Clara Foods, Modern Meadow, and C16 Biosciences.

Customers pay a monthly rate based on the number of bioreactors they use and the scale of the data they need. Unlike some development partners, Culture doesn’t get royalties when products are commercialized.

“People really like predictable costs and guaranteed capacity,” Patrick says. “We think it’s a business model that works well for us and our customers.”

A neuroscientist and a cancer biologist walk into a room.

It sounds like the start of a nerdy joke, but it was the scene at Cold Spring Harbor Laboratory when a group of scientists gathered to discuss how conversation between nerves and tumor cells might help cancers grow. Convened by a biotech start-up called Cygnal Therapeutics, the group discussed not just how the central nervous system might influence cancers in the brain and spinal cord but also how the peripheral nervous system—the bundles of nerves that control everything else—might be taken advantage of by cancers.

Credit: Cygnal Therapeutics

Pearl Huang, CEO of Cygnal Therapeutics

By the end of the December 2019 meeting, the scientists had coined a phrase for this fairly uncharted territory: cancer neuroscience (Cell 2020, DOI: 10.1016/j.cell.2020.03.034). And Cygnal, hoping with the meeting to build both community and consensus around the idea that the peripheral nervous system could influence nonnerve diseases, has since broadened the concept beyond cancer to a host of diseases. Calling what they study exoneural biology, the scientists at Cygnal see the relationship between peripheral nerves and the diseased cells around them as an area ripe with targets for therapeutics.

CEO Pearl Huang says what they are doing has immensely broad scope.

“Everybody else is out there with new modalities, new widgets, new ways of going back to old targets,” says Huang, a seasoned pharma and biotech leader who marshaled teams at Roche, GlaxoSmithKline, and Merck & Co. before coming to Cygnal. “We’re opening up a whole new field of human physiology.”

In many cancers, diseased cells talk to blood vessels, recruiting them to snake around and through the tumor to provide it with nourishment. Huang and others think diseased cells also have something to gain from neurons. Certain cancers are dense with nerves, and in some cases, like with stomach cancer, removing those nerves slows tumor growth (Sci. Transl. Med. 2014, DOI: 10.1126/scitranslmed.3009569).

A nervy approach to disease

Credit: Yang H. Ku/C&EN/Shutterstock

Cygnal Therapeutics believes that peripheral nerves can influence disease. With evidence that diseased cells can coax neurons to grow toward them, the company uses a combination of cell culture experiments, genome screening, bioinformatics, and other techniques to discover drug targets and small molecules that can disrupt the communication between diseased cells and the nerves nearby.

What exactly happens at a molecular level between these nerves and cancer cells is still a bit of a mystery, Huang says, and among the young biotech’s first tasks is trying to decipher some of these conversations.

Cygnal has already found some clues. Since it was spun out of the life sciences venture capital firm Flagship Pioneering in late 2019, the company has begun to develop two compounds. One is a small molecule that modulates a target that aids nerve synapses and plays a role in the development of cancer. The other is a large molecule that modulates a protein suspected of helping move neurons to and away from tumors.

Cygnal’s scientists are using several methods to interpret the conversations between neurons and the healthy and diseased cells around them. They include cell culture, animal disease models, microscopy, and chemical techniques that affect neuronal function.

The scientists focus on the 2,000 genes linked to neuronal function and use bioinformatics to try to determine what those genes are doing when neurons are talking to the nonneurons nearby. They might be looking for a gene that screams in disease and whispers in health, or an entire signaling cascade that chatters loudly in health and goes all but silent in disease. The team also uses CRISPR to see what genes are vital to this talk and screens drug libraries for small molecules that can interrupt these conversations. From there, Huang says, the scientists can identify molecules that could be developed into drugs.

Some targets unearthed by the approach are associated with brain diseases, so it’s not too far a leap to wonder if these targets might affect diseases influenced by peripheral nerves. That has given the team confidence it’s on the right track.

“I think exoneural biology is relevant to every disease,” Huang says. “And there is so much work to be done.”

Credit: Evrnu

Christopher Stanev (left) and Stacy Flynn, founders of Evrnu

There comes a day when you must part ways with your beloved old Pink Floyd T-shirt—splattered with paint, perpetually wrinkled, and too torn to be worn outside on a chilly day. Into the trash it goes, right?

Wrong. Or at least it could be, according to the founders of Evrnu, a start-up looking to use its technology to give textiles a second life—and a third and a fourth.

Stacy Flynn, Evrnu’s CEO and cofounder, spent much of her career in the garment industry, working as a global textile specialist for companies like DuPont and Eddie Bauer. Several years ago on a trip to an area of China concentrated with textile subcontractors, she and a colleague realized they couldn’t see each other’s faces through the smog. Pollution from the textile process clogged the air and blackened the water.

“I came out of that 1-month trip thinking that this cannot be how the story ends,” Flynn says.

Later, when working as a fabric specialist for Target, Flynn met Christopher Stanev, now the president and cofounder of Evrnu. Stanev had spent many years working in textile engineering and would later hold technical leadership positions at companies like Nike and Gloria Jeans. They dreamed up and started Evrnu, a company that creates and licenses technologies to allow textiles to be continually recycled rather than tossed into landfills.

Nu life for old clothes

Credit: Adapted from Evrnu/Yang H. Ku/C&EN/Shutterstock

Evrnu’s patented NuCycl technology allows cotton waste to be dissolved and extruded into cellulose fibers. Those fibers can then be respun into yarn, which in turn can be woven to make textiles for new garments.

Evrnu developed its first technology in-house and launched it in 2019. Called NuCycl, it’s a recycling process in which cotton garments are stripped of dyes, prints, and other adornments, leaving behind only cellulose. The cellulose is purified and activated into a pulp that is ready to be dissolved in a solvent. Once the pulp is liquefied, cellulose-based fibers can be extruded from it and used to make new fabric.

The fibers produced through Evrnu’s process, and by other textile recycling start-ups such as Renewcell and SaXcell, are similar to or the same as human-made cellulosic fibers like viscose, acetate, and lyocell. But most of those human-made fibers are produced from wood pulp, which is lower in cellulose content than pulp made from cotton garments. Rather than make cellulosic fiber out of harvested trees, the start-ups argue, it’s better to make it out of the discarded cotton clothes that stack up by the millions of pounds.

Evrnu sees that waste as a valuable resource for producing high-quality cellulose-based textiles. And because the cotton pulp is higher quality than wood pulp, “the ending fibers are stronger, more elastic, and can be worn a longer time,” Stanev says.

Evrnu has licensed NuCycl technology to brands like Adidas, which has used it to produce the Stella McCartney Infinite Hoodie. It’s a sleek, cream-colored sweatshirt intended to be returned to Adidas and recycled indefinitely. Its fibers are so high quality, Evrnu says, that the garment can easily be put through the NuCycl process again and again and turned into brand-new apparel without compromising on quality.

And Evrnu’s team doesn’t plan to stop there. Among other projects, it’s working on further innovating the solvent in which the cellulose is dissolved to produce fibers. The firm says it is using “a blend of different solvents and catalysts” that can dissolve longer molecules, allowing the polymer chains to maintain their integrity through multiple cycles of reuse. The company is also looking to expand its textile recycling technology to materials other than cotton.

Credit: Lilac Solutions

From left: Lilac Solutions' chief development officer, Tom Wilson; chief operating officer, Nick Goldberg; and CEO, David Snydacker, at the firm's Oakland, California, facility

When David Snydacker began his PhD in materials science at Northwestern University, his goal was to find new materials that would soup up the performance of electric-vehicle batteries. After 5 years, he realized that companies like Tesla were doing just fine and probably didn’t need his help.

But even as battery technology improved, lithium remained the common denominator. And all the lithium that ends up in batteries is pulled from the ground with the same two resource-intensive methods. So, in the final year of his program, Snydacker shifted gears and began applying the tools of battery materials discovery to developing a new way to extract lithium.

In June 2016—the same month he graduated—Snydacker formed Lilac Solutions. This February, he was rewarded for his progress with $20 million in series A funding from investors led by Breakthrough Energy Ventures, the Bill Gates–backed venture capital fund.

Snydacker convinced his investors that today’s methods of obtaining lithium—hard-rock mining and huge brine evaporation ponds—aren’t going to cut it if the world is to shift wholesale to electric vehicles. His alternative was the ion exchange method the Lilac team had spent the previous 4 years honing.

The briny deep

Credit: Adapted from Lilac Solutions/Yang H. Ku/C&EN/Shutterstock

Most lithium is recovered by sucking lithium-rich brines out of the ground and into large ponds that are left to evaporate for months or even years to concentrate the lithium (left). Instead, Lilac Solutions taps directly into the underground brine, running it over its ion exchange material to concentrate the lithium in real time (right).

Snydacker landed on the approach after considering—and rejecting—other lithium-refining techniques, including electrochemistry and solvent extraction. Ion exchange is already widely used to remove ionic impurities from water by swapping a target ion, like calcium, with another one, like hydrogen. His challenge was finding an exchange material that could stand up to the hydrochloric acid needed to flush out the accumulated lithium as lithium chloride. “No conventional ion exchange materials were suitable for this job,” he says.

To start his hunt, Snydacker searched the Open Quantum Materials Database, a library of more than 600,000 materials created by his Northwestern adviser, Chris Wolverton. He found 400 lithium metal oxides that could potentially do the job. After launching Lilac, Snydacker broadened the search and, with the help of a Small Business Innovation Research grant from the US Department of Energy, found 1,000 more potential materials. “We expanded and looked at the whole periodic table,” he says.

Lilac’s scientists then began to narrow down the list at their labs in Oakland, California. In 2017, after many months of testing, they settled on what Snydacker calls “a ceramic with a unique composition and crystal structure.” They coated it with a thin inorganic oxide, similar to what coats battery cathodes, to help keep it stable.

Since then, Lilac has tested the ion exchange material on truckloads of lithium-containing brines brought to Oakland from sites in the western US and South America. Snydacker says the firm and partner companies will soon move the test equipment to the field in both regions. Companies that ultimately adopt Lilac’s process will use conventional means to convert the lithium chloride into battery ingredients like lithium carbonate and lithium hydroxide.

Snydacker knows that the path from these pilot projects to commercialization will not be easy or cheap. “The first full-scale lithium project will be hundreds of millions of dollars,” he says.

The partners that Lilac has disclosed so far are small mining ventures with big plans but not a lot of money. Snydacker says his firm is also working with much larger companies that it hasn’t disclosed. Regardless of its size, any partner will require detailed feasibility studies before giving a project the green light.

The market research firm Roskill predicts that demand for lithium compounds will grow more than 18% per year through 2030—a pace that could ramp up if more places, like California, set a date for banning gasoline-powered vehicles. That demand will necessitate much new production capacity. A brine evaporation and processing plant of the type now found in Nevada and South America, Snydacker says, takes a good 10 years to get off the ground. “We’re able to reduce that to about 2 years,” he says.

Credit: Lycia Therapeutics (Trombley), Laura Morton Photography (Bertozzi)

Lycia Therapeutics CEO Aetna Wun Trombley (left) and founder Carolyn Bertozzi

It’s an open secret in the drug industry: pinpointing the genetic cause of a disease is increasingly easy; fixing that disease is still hard. That’s because conventional drugs are able to attack only a small fraction of the misbehaving proteins that can wreak havoc in the body

Lycia Therapeutics is part of a wave of companies developing a technology called protein degradation to expand the accessible fraction of the proteome. Instead of blocking the activity of a protein, molecules known as degraders cause a protein to be broken down by our body’s waste disposal system.

In recent years, big and small firms alike have devoted significant time and energy to developing protein degraders. Most of those efforts have used complex small molecules to target proteins inside cells, aiming to send them to the proteasome, one of the body’s trash bins. In contrast, Lycia targets the roughly 40% of proteins that live and work outside the cell. The start-up tags them to be broken up in an enzyme-packed organelle called the lysosome.

Break it down

Lycia Therapeutics is developing LYTACs (or lysosome-targeting chimeras) that force the degradation of extracellular or transmembrane proteins. It designs molecules—in this case an antibody—that bind to both a bad-behaving protein and a receptor that pulls the construct inside the cell, where it is directed to the lysosome for breakdown.

Credit: Adapted from Nature/Yang H. Ku/C&EN/Shutterstock

“Protein degradation has been a hot field, but what has been elusive is how do you apply this to the other half of the proteome,” says Lycia’s CEO, Aetna Wun Trombley. Drug developers would love to take down not only circulating proteins but also ones that have aggregated and immune complexes that have gone awry.

Stanford University chemical biologist Carolyn Bertozzi came up with a breathtakingly simple solution: design an antibody that can bind to both a protein target and a receptor that would drag the protein into the lysosome, where it would be digested. She called the construct a LYTAC, for lysosome-targeting chimera.

Bertozzi’s lab started with an antibody that binds to apolipoprotein E4, which is linked to Alzheimer’s disease. The researchers then engineered the antibody to express a ligand for a well-studied glycoprotein called cation-independent mannose-6-phosphate receptor that connects cell surfaces to the lysosome.

To show that the concept could be generalized, researchers in Bertozzi’s lab created LYTACs targeting a handful of other proteins that are high on drug developers’ wish lists, such as the cancer immunology target PD-L1, against which there are already-approved antibody therapies. In the lab at least, the molecules led to the protein targets’ being crunched down by the lysosome.

In March 2019, Bertozzi posted the paper on the preprint server ChemRxiv, meaning it wasn’t peer-reviewed. To her surprise, it immediately attracted attention. She found herself fielding queries from companies and investors interested in pushing the concept further. Among them was Versant Ventures, which in short order licensed the technology and had a team of scientists at its labs in San Diego work on validating and expanding on those original tool compounds.

Fast-forward to June 2020, when Lycia formally launched with $50 million from Versant, and Trombley at the helm. The following month, Bertozzi’s paper was published in Nature (2020, DOI: 10.1038/s41586-020-2545-9). Lycia has since moved out of the San Diego incubator to its own space in South San Francisco, where it expects to staff roughly 30 researchers by the end of the year.

First on the docket for Trombley is setting a research strategy for Lycia. She says the firm’s scientists are exploring targets that are the subject of clinical-stage drug programs or even marketed drugs, as well as previously intractable proteins. In either case, the biology needs to be really clear, or clinical data for a LYTAC will be difficult to parse. Without clear biology or clinical data, “if it’s not working, you don’t know if it’s because of the target or because what you designed didn’t work,” she says.

The Lycia team has also broadened the company’s scope beyond Bertozzi’s original construct. “She started working with antibodies, but you can think about the flexibility these types of molecules can afford,” Trombley says. That includes tinkering with both sides of the construct: identifying ligands against other receptors that can route a drug to the lysosome and swapping out the antibody for, say, a small molecule.

Credit: Natron Energy

Colin Wessells, CEO and founder of Natron Energy

The world’s thirst for power has pushed engineers to think hard about how we use and store energy. We’re familiar with lithium-ion batteries in our cell phones and laptops, but less-visible, stationary batteries—often used for emergency power—rely on older technologies, like lead acid. Both technologies pose safety, environmental, and performance problems. Natron Energy believes its Prussian blue battery is a safer and more reliable option.

In 2012, Colin Wessells was finishing his PhD in materials science at Stanford University and working on electrode materials. Robert Huggins, a professor he was working with, recalled old research on Prussian blue—a pigment created by the oxidation of ferrous ferrocyanide salts—showing it could make windows change color with the application of an electric charge. Huggins suggested it could also be used to store and release energy. Wessells was tasked with figuring out “What is battery-grade Prussian blue, and how do we adapt this material system for energy storage?”

Wessells delved into research and discovered that Prussian blue’s crystal structure allows for an almost friction-free transfer of electrons, meaning a Prussian blue–based battery can charge and discharge significantly faster than a lithium-ion battery without risk of explosion. He realized he had something with commercial potential, so as soon as he finished his PhD, he launched Natron.

Safer, faster, longer

Credit: Adapted from Natron Energy/Yang H. Ku/C&EN/Shutterstock

Natron Energy uses a Prussian blue analog for the anode and cathode, which contain crystalline nanoparticles of manganese hexacyanomanganate. The cubic open-framework structure—made of iron and manganese (small spheres) bridged by cyanide ligands (bars)—allows sodium ions to pass through with almost zero strain. This ultralow resistance allows quick energy discharge and long cycle life.

Natron has developed a battery made of low-cost materials that can discharge in 2 min or less and charge in 8 min. Made primarily of salt, iron, and manganese, it is touted as having a smaller ecological footprint than both lithium-ion batteries, which may use rare-earth metals, and lead-acid batteries, which contain large amounts of the toxic metal lead.

But Prussian blue batteries aren’t as energy dense as lithium-ion batteries and can’t replace them in most uses. Natron does not see that as a drawback. “There’s no such thing as a perfect battery; there are appropriate batteries for every situation,” says Natron’s vice president of sales, Jack Pouchet. Natron has aimed its battery technology at data centers, electric forklifts, and small electrical grids—applications that demand a quick discharge of power.

As more electric vehicles hit the road and many pull into charging stations at the same time, the sudden increase in power demand can strain electrical grids. Natron figures its battery, installed at a charging station, can address that problem by providing the initial surge of power needed to supercharge electric vehicles.

Prussian blue batteries can also replace lead-acid batteries typically used for backup power at data centers. In a power outage, Natron’s batteries can provide a quick surge before generators kick in.

Natron says its batteries are long lasting. While lead-acid batteries have a service life of 1–2 years, Natron says its batteries perform for 5 years or more and have lasted 35,000 charge-discharge cycles without degradation. The company is confident that more than 50,000 cycles are possible.

This reliability has proved attractive to investors. In September, Natron was awarded $20 million from the US Department of Energy’s Advanced Research Projects Agency–Energy. With this new funding and $35 million in series D funding raised earlier this year, Natron plans to scale up battery production. It’s spending some of the money to firm up its supply chain and continue research to increase the battery’s energy density.

The company recently received a UL 1973 listing for its BlueTray 4000 rack-mounted battery, certifying its safety for use in stationary applications. The listing “allows us to start selling batteries generally” for commercial purposes rather than just research, Wessells says. “We are now shipping battery systems as fast as we can,” he says, “and we’re overbooked with customers.”

Natron has about 57 employees at its Santa Clara, California, headquarters. It expects to more than double that number as it builds new production centers, including one in Valais, Switzerland, funded by a grant from the Swiss government.

Credit: Protera

Protera founders Francia Navarrete (left) and Leonardo Álvarez in their R&D lab in Chile

Protera started with an ambitious goal: use artificial intelligence to rationally design functional enzymes and other proteins. Leonardo Álvarez and Francia Navarrete had seen computational methods used to validate and interpret results in the biotech world, such as Insilico Medicine’s drug-candidate finder, and they wondered if they could use AI earlier in product development.

They founded Protera in 2015 in Chile, where they both grew up and went to school. In 2017, the company joined the IndieBio accelerator in San Francisco. In June of this year, Protera received $5.6 million in a series A investment round led by Sofinnova Partners. As part of that deal, Protera is scaling up operations with partners in France. The business office will remain in Silicon Valley, while R&D is in Chile.

The international footprint came about for business reasons, but it also exposes Álvarez and Navarrete to a wide range of potential partners and product ideas. San Francisco’s foodie culture inspired their first product, Protera Sense, an enzyme that can convert unsaturated fats into saturated fats, which have higher melting points—for example, turning sunflower oil into a buttery spread.

Protera’s enzyme lets food makers create textures they would normally get from palm oil or trans fats, two things that eco- and health-conscious consumers are becoming less willing to tolerate. “The next-generation food consumers are deeply involved with the ingredients that the products they’re eating contain. They want to know that these ingredients are healthy, sustainable, and clean label,” Álvarez says.

Protein lingo

Protera's Madi artificial intelligence system marries linguistic computational methods with deep learning and other approaches to find connections between protein function, sequence, and structure. Protera scientists can input a reaction they'd like to catalyze, and the AI will yield DNA sequences encoding a protein that should do the job. From there, they use bacteria to make test samples of the protein and then scale up the winners in yeast.

Credit: Yang H. Ku/C&EN/Shutterstock

Clean label is a nebulous but popular term that means a product, usually consumable, is made from naturally derived ingredients rather than synthetic ones. It’s a slightly less judgmental version of the “chemical-free” marketing trend that can rile chemists up. Ingredients made by fermentation are generally considered clean label. Protera expects that those made with its functional proteins will also fit the bill.

Álvarez is bullish on what his system can do. “We know that proteins are functional; you can virtually do everything with proteins. But you do need to understand how to optimize them,” he says. “You need to understand how they work. So that’s where the AI platform will play a key role.”

The computational core of Protera’s approach is an AI system that maps correlations between the sequences, structures, and functions of proteins. The system mines those correlated data sets to predict the structure a certain amino acid sequence will adopt or suggest an amino acid sequence to yield a desired structure. It can even start with a desired reaction and yield DNA sequences for proteins to do the job.

Going from sequence to structure has a lot in common with translating English or Spanish to a symbolic language such as Chinese, Álvarez says. In fact, Protera’s senior AI engineer, Roberto Ibáñez, has a background in applying deep learning and neural networks to language translation and text generation. Working with protein geometries requires other, less-well-trod types of algorithms.

Protera runs its AI on specialized graphical processing units, allowing it to perform a large number of calculations in parallel. Nvidia, a Protera investor, first designed such chips for video games, Álvarez says, but the AI world has enthusiastically adopted them.

Once the computer puts out DNA sequences, Protera uses normal biotech tools and fermentation to screen and produce the protein.

Protera’s next product, which is going into pilot-scale fermentation now, is a thermally stable antifungal called Protera Guard. The company is scaling up production of the protein and working with French bakeries to test it. Early results suggest the protein preservative can extend the shelf life of breads by up to 45 days.

“The market for clean-label ingredients is growing superfast,” Álvarez says. It’s expected to reach $64 billion by 2026, he says, with clean-label food preservatives alone growing to about $2.5 billion. Meanwhile, the market for edible palm oil replacements is close to $3 billion. Given such strong demand, Protera expects to have plenty of field to play on with its products.

Credit: Ventus Therapeutics

Top row, from left: Ventus Therapeutics CEO Marcelo Bigal, cofounder Hao Wu, and cofounder Judy Lieberman. Bottom row, from left: cofounder Richard Flavell, cofounder Thomas Tuschl, and Ventus head of R&D Michael Crackower

The innate immune system is our first line of defense against microbial invaders, and a protein called NLRP3 is one of its loudest alarms. When triggered by an intruder, NLRP3 proteins change shape and begin forming a large structure called the inflammasome, which in turn activates a flurry of molecules that beckon the immune system to send backup.

But it doesn’t stop there. The inflammasome also initiates a cellular self-destruct sequence in which proteins called gasdermins kill the cell by poking holes in its membrane from the inside out.

Our bodies go to great lengths to protect us from infections. Maybe too great.

Such sensitive alarm systems can be a liability. Over the past decade, scientists have linked hyperactive NLRP3 inflammasomes to a dizzying list of diseases, including Alzheimer’s disease, atherosclerosis, inflammatory bowel disease, nonalcoholic steatohepatitis (NASH), and Parkinson’s disease.

Stopping the self-destruct sequence

Credit: Adapted from Hao Wu/RCSB Protein Data Bank ID 6NPY/Yang H. Ku/C&EN

Ventus Therapeutics is designing small-molecule drug candidates that target proteins of the innate immune system, including NLRP3. When individual NLRP3 proteins are activated, they aggregate to form a wheel-like oligomer, which is part of the NLRP3 inflammasome. That inflammasome helps activate and release inflammatory molecules. It also causes the formation of other protein structures, called gasdermins, which poke holes in a cell and kill it. This self-destruction system is a last-ditch effort to rid a cell of microbial invaders. But improperly activated NLRP3 inflammasomes can also lead to a bevy of diseases. Ventus is designing compounds to stop inflammasome activation.

With a record like that, it’s no surprise that NLRP3 is moving up on pharma’s most wanted list. In the past 2 years, Bristol Myers Squibb, Genentech, Novartis, and Roche have all acquired start-ups developing small molecules that take aim at NLRP3. And in May, a new start-up called Ventus Therapeutics formally debuted with $60 million and a fresh strategy for targeting NLRP3.

It’s actually the second start-up backed by Versant Ventures to target NLRP3. In 2018, Genentech acquired the Versant-funded Jecure Therapeutics, which was developing small-molecule NLRP3 inhibitors to treat NASH. Jerel Davis, a managing director at Versant, thought there was plenty of room for another inflammasome company, and Michael Crackower, who is now head of R&D at Ventus, was charged with figuring out how to make the sequel even better than the original.

The new firm would take a broader look at innate immunity by investigating NLRP3, other inflammasome proteins, and different innate immune system alarm bells, such as the cGAS-STING pathway, which detects viral and cancer DNA. But Crackower wanted to set Ventus apart not just in its breadth of targets and potential disease areas but also in how it discovered its drug candidates in the first place.

Jecure had relied on phenotypic screens in which researchers added different compounds to cells in hopes of finding one that quieted NLRP3’s blaring alarms. But that strategy left scientists wondering what the molecules were doing to NLRP3 or if they were even targeting the protein directly. “They were really limited in how they could tackle these targets,” Crackower says.

Crackower wanted to take a more direct approach. The problem is NLRP3’s natural propensity to shift its shape and clump together. That feature is key for quickly forming active inflammasomes in sick cells, but it causes headaches for researchers trying to study the proteins in a test tube. Those clumps make it impossible for scientists to screen drugs on the protein itself or capture a clean picture of its molecular structure.

Hao Wu’s lab at Harvard Medical School found a solution through protein engineering. By changing amino acids at key interfaces, forming chimeric proteins, and purifying proteins in just the right way, her group developed methods to effectively paralyze the shape of NLRP3, gasdermin proteins, and other key players in the innate immune system and solve their structures.

Wu realized that her lab’s techniques could help drug companies design compounds that specifically target the structure of NLRP3, as well as related proteins. She founded a small company called Smoc Therapeutics to explore the idea. Last year, Ventus acquired Smoc’s assets and recruited Wu and several other innate immune system experts to be the scientific cofounders of Ventus.

Ventus CEO Marcelo Bigal thinks the firm’s structure-based drug discovery approach to NLRP3 and other innate immune system targets will give it an advantage over more advanced and well-funded competitors. Other companies discovered NLRP3 inhibitors by “trial and error,” he says. Ventus is doing “drug discovery with the lights on.”

Credit: Doug Levy/the Engine

From left: Zapata Computing founders Peter Johnson, Christopher Savoie, Jonathan P. Olson, Alán Aspuru-Guzik, Jhonathan Romero Fontalvo, and Yudong Cao

When computational chemist Alán Aspuru-Guzik told his business partner Christopher Savoie that he had an idea for a quantum computing software company, Savoie was skeptical, he admits.

This was 2017, and quantum computing was (and arguably still is) in its infancy. Savoie asked Aspuru-Guzik what a state-of-the-art quantum computer could do. Not much more than a pocket calculator, Aspuru-Guzik told him.

But the two had worked together successfully at Kyulux on machine-learning methods to advance organic light-emitting diode technology, so Savoie visited Aspuru-Guzik’s lab, then at Harvard University. By the time they’d finished lunch, Savoie says, he was convinced that Aspuru-Guzik and his collaborators—Yudong Cao, Peter Johnson, Jonathan P. Olson, and Jhonathan Romero Fontalvo—were on to something. By the afternoon, Aspuru-Guzik was calling quantum computing researchers to introduce Savoie as the new CEO of Zapata Computing.

Quantum computing proponents have made huge promises, especially in chemistry. A traditional computer represents information as ones and zeros. But quantum bits, or qubits, can represent a one and a zero simultaneously. That could allow quantum computers to more realistically represent the quantum weirdness that underlies all molecular behavior. The computational chemistry methods that scientists currently use to simulate the behavior of drug molecules or specialty chemicals tend to be fast or accurate but not both. Quantum computers might change that.

Preparing for the leap to quantum

Credit: Yang H. Ku/C&EN/Shutterstock

The bits in your computer represent data as ones and zeros, but a quantum computer's bits can be a one and zero simultaneously. That's why some experts think quantum computers could make computational chemistry simulations more accurate. But it could be decades until scientists iron out all the kinks in these cutting-edge machines. In the meantime, Zapata Computing offers companies expertise and software to help them plan for the future. Zapata helps companies evaluate the quantum computing options that are available—such as those from companies D-Wave Systems and Rigetti Computing—and test and refine chemistry simulations on those machines.

With hype swirling about a quantum computing revolution, it’s no surprise that industrial companies like BASF and Robert Bosch—both Zapata investors—want to get in on the action. The question is: How? Several firms, including IBM, Google, and D-Wave Systems, have demonstrated quantum computers, though these still have only a few dozen bits and are prone to errors. Each uses slightly different technology, posing a problem for businesses deciding which horse to back.

Savoie says companies see that they need to get software development plans in place for quantum computing. They can’t wait for machines with hundreds of qubits and all the kinks worked out.

“You can’t just wait until the day a 300-qubit computer shows up and expect to have everything in place and hit go,” Savoie says. “You need to start developing those things earlier.”

Zapata, which takes its name from the Mexican revolutionary Emiliano Zapata, advertises a way in. In 2013, Aspuru-Guzik, now at the University of Toronto, invented the first algorithm that could make accurate predictions even on imperfect qubits. Back then, his group was also figuring out how to make that type of algorithm work on different quantum computing hardware and learning the strengths and weaknesses of those computers.

Now Zapata offers its algorithms and expertise to customers. It can advise a company on where to make investments related to quantum computing, help it determine how to harness quantum computing, or set up comparisons of different quantum computing platforms. Savoie says Zapata can even suggest ways to avoid quantum computing; some problems might be solvable with mostly classical computing.

In September, Zapata released a software platform called Orquestra. It lets users create and adjust workflows for quantum computers and then test and optimize those processes on different machines.

Savoie says there’s no way to know when large, error-free quantum computers will be available for commercial use. It could be a decade, or two, or never. Whatever happens, he says, companies can rely on Zapata’s software and expertise to guide them.