Men and woman in white lab coats and purple rubber gloves stand beside metal tables, looking at small glass-enclosed cabinets. Inside, a robotic arm carefully lowers a clutch of what look like small eyedroppers to a specialized tray containing small plastic reservoirs, deploying a careful measure of liquid into each. Sensors and microchips embedded in the tray detect and automatically analyze the resulting chemical reaction, sending a stream of data to nearby computer. The atmosphere is hushed and serious, but I keep expecting the bald visage of Dr. Bunsen Honeydew and his hapless red-haired assistant Beaker—the mad scientist characters from The Muppets—to pop up from behind a lab bench.

Instead, I’m being led through the lab by James Field, who does look every bit the part of the mad scientist, particularly in his lab coat. Field is thin, with a Young Einstein–like whoosh of curly brown hair piled on his head, and, on the day I meet him, is wearing round, tortoiseshell glasses. He is the founder and chief executive of LabGenius, the London startup that owns this high-tech robotic lab, located in a former cookie factory.

The company, which Field founded in 2012 when he was still completing his Ph.D. in synthetic biology at London’s Imperial College, is combining advances in robotic chemistry and artificial intelligence to create new proteins that could be the basis for important new medicines.

Today, LabGenius announced it has received an additional $15 million in venture capital financing led by Atomico, the London-based venture capital firm created by Skype cofounder and billionaire Niklas Zennström, to help it to further grow its business.

Field says the funding will help LabGenius in numerous ways. “It’s more robots, more experiments, more people to run the robots,” he notes, but also to develop more sophisticated automated lab tests to enable the company to make more complex proteins.

The valuation terms of the new financing were not disclosed, but documents filed with U.K. business registry Companies House indicate that new funding valued the company at about $53 million.

Atomico partner Irina Haivas will join LabGenius’s board as part of the investment deal. Previous investors in LabGenius, including Lux Capital, Obvious Ventures, Kindred Capital, and Inovia Capital, are also participating in the latest financing.

Haivas, the Atomico partner, says she sees LabGenius helping to solve a “productivity crisis” in the pharmaceutical industry, with the rate of new medicines being discovered slowing and the expense of each new drug brought to market rising. “The way they work hasn’t changed in more than 10 years,” she says, noting that most drug companies still find new medicines from a human-driven hypothesis that is tested through manual, human-run experiments, often producing data that is difficult to capture and verify. “I don’t see a way to get out of this—without supporting the human with technology and data,” she says.

The vast majority of today’s medicines are made using small molecules. But proteins—which are much larger and more structurally complex substances—are the basis of an emerging class of therapies often referred to in the pharmaceutical industry as “biologics.” These are lab-engineered proteins that can mimic the natural antibodies and enzymes that regulate many functions within the human body. Currently, protein-based therapies make up about 30% of all drugs. They include treatments for many cancers, autoimmune disorders, and rare diseases. Protein-based therapies are often far more targeted, and have fewer harmful side effects, than drugs based on small molecules, but they are considered more difficult and costly to develop.

The problem, as Field explains, is that there are so many possible DNA combinations that create proteins that it is extremely difficult to figure out exactly which combination might yield an effective protein therapy. Using traditional methods, designing a single, useful protein can take a researcher years. LabGenius is using robotic automation, synthetic biology, and machine learning to speed that process up.

LabGenius has set its sights on inflammatory bowel disease as the first disease it wants to target. Field says the area is promising because several drugs for the condition already exist, but they have to be injected into patients because they are not chemically stable enough to be taken orally and then survive the harsh environment of the human digestive tract to reach the bowel. So LabGenius has been working to create proteins that would perform similarly but could be put into a pill. The company has already produced several possible proteins and has shown in its lab that they can survive digestive enzymes and the acidic environment found in the stomach. The next step, he says, is conducting safety trials with animals.

The company uses its robotic lab, which can make and test proteins much faster and more reliably than experiments performed by humans, to create a library of information about the biological properties of proteins it is designing, such as how stable they are or how much they react with certain other substances that are of clinical interest. Then the company uses this library to train a machine-learning model to predict how certain DNA sequences correspond to a protein’s properties. In this way, it can drive the proteins it is designing toward the qualities that may make them useful drugs.

“They are integrating existing equipment with the biological methods and A.I. and their data set to get this flywheel effect,” Haivas says.

The robotic lab and machine-learning process is similar to that used by Zymergen, an Emeryville, Calif., company that is focused on creating genetically modified crops and new kinds of biologically engineered materials for the electronics and consumer care markets. It was valued at $874 million after a funding round in September. Aaron Kimball, Zymergen’s chief technology officer, is an adviser to LabGenius.

It’s also similar to the method used by Recursion Pharma, a Salt Lake City–based company, now valued at $465 million. It uses a high-speed robotic lab to screen the effect of small molecules on cells.

The difference, Field says, is that large libraries of small molecules and their properties already existed for Recursion to use. For proteins, the DNA libraries are not nearly as comprehensive, he says, requiring LabGenius to gradually build up its own. (Chris Gibson, Recursion’s cofounder and chief executive, is an investor in LabGenius too.)

Because of its potential impact in drug discovery, predicting the structure of proteins has also been a focus for artificial intelligence researchers, including teams at both DeepMind, the London-based A.I. company owned by Google parent Alphabet, and Facebook. Two years ago, DeepMind created an algorithm called AlphaFold that used a deep neural network—a kind of machine learning loosely based on how the human brain works—to trounce the competition at a closely watched biennial competition for protein structure prediction. But even then, AlphaFold could only correctly predict a protein’s structure about 55% of the time.

Field says that by learning the correspondence between a DNA sequence and the biological properties of the protein it encodes, LabGenius can skip the difficult step of having to predict the structure of the proteins it is designing. Knowing function, he says, is more important than knowing form. After all, he says, Darwinian evolution doesn’t have any understanding of the structures it is creating—it simply has a functional imperative. “Why argue with nature?” Field says.