Since the theatrical release of Jurassic Park in 1993, the thought of reviving long-extinct creatures has been a part of popular culture. And with the help of scientific advancements in the 21st century, the idea has steadily come closer to reality — like a project in Texas that’s using genetic engineering and cloning to try to bring back the dodo and wooly mammoth.
In a future where we could recreate species in the lab, it doesn’t necessarily mean we should. There are ethical and philosophical concerns with de-extinction (such as, would we resurrect beings just to isolate them in zoos? Or send them to a second extinction in environments now ravaged by climate change?), along with the practical concerns of, say, a snarling T-Rex visiting you on the toilet.
But what if instead of recreating the entire genetic code of fossilized mammals or birds, we could bring back something smaller, like an individual molecule or a genetic mutation, from a long-ago organism? And what if that could save millions of lives every year?
One Penn biologist is working to achieve that goal, having resurrected numerous molecules from ancient organisms which have shown potential in fighting infections today — including a groundbreaking discovery in Neanderthals. “This is the first molecule discovered in an extinct organism that has therapeutic potential,” says César de la Fuente, who runs the de la Fuente Lab and the Machine Biology Group at Penn. “We brought it back to life through chemistry in the laboratory.”
De la Fuente coined the term “molecular de-extinction” years ago to refer to the novel work of his lab, which now houses 20-plus researchers on Penn Medicine’s campus. Their transdisciplinary methods utilize artificial intelligence and machine learning to locate new candidates for antibiotic research, including molecules that were once buried in permafrost for thousands of years. “It’s a state of the art, AI model that we trained from scratch,” says de la Fuente, “we call it APEX, and it enabled us to mine every single extinct organism known to science, all throughout evolutionary history. This AI model has opened a window into the past to find potential solutions to present-day problems like antimicrobial resistance.” Though the team’s research is still years away from clinical studies, their experimental validations have sparked excitement, receiving coverage from NPR, CNN, and Nature, and others.
“All the science that we’re doing using AI and these different tools, hopefully someday it helps improve the world by saving lives.” — César de la Fuente
The lab’s work is part of a critical push in modern medicine for fresh antibiotics, one that’s desperately in need of out-of-the-box ideas. Since the 1980s, the scientific community has not developed a new class of antibiotics, despite the escalating need for discoveries. Around the world, more than 7 million people died from bacterial infections last year, making them the second-leading cause of death globally. What’s more troubling from a public health standpoint is that two-thirds of those deaths are a result of antimicrobial resistance, or AMR, a result of bacteria developing immunities to the drugs we use to treat them.
“I think [de la Fuente’s] work is incredibly interesting and important, especially with the looming threat of antimicrobial resistance in bacterial pathogens,” says Dr. Seth Shipman, an associate professor in the Department of Bioengineering and Therapeutic Sciences at the University of California-San Francisco and an investigator at the nonprofit Gladstone Institutes.
“We can’t keep making one-step derivatives of the current drugs. We need creative strategies. Otherwise, we’re in real existential danger.”
From Marine Biology to Machine Biology
Growing up in coastal Spain, de la Fuente, 38, spent a lot of time training for his future profession. But he was just being a kid. He dissected samples of marine life. He studied how geckos regenerate their tails. One Christmas, after meticulous calculations, he asked for enough helium balloons to lift himself up and fly — a request which Santa Claus denied.
“I think we’re all born scientists, constantly wondering about how the world works,” he says. “Unfortunately, as we age, we tend to lose that curiosity. Society kind of kicks it out of us.”
That childlike sense of wonder extended to computers and science fiction as well. De la Fuente describes an adolescent fascination with “how machines can help amplify what makes us human,” recalling the influence of films like Blade Runner and The Matrix on his imagination. Later, after earning his doctorate, de la Fuente began musing on how to accelerate our investigations of nature with computers, and even how they might play a role in “programming” biology.
Since the beginning of antibiotic research in the late 19th century, the field has relied on painstaking work. Antibiotics are a class of medicines that derive from living things in the world. Traditionally, they’ve been found in soil, water, or organic matter. But instead of spending years collecting samples in the wild and studying them, what if there was a shortcut?
Unlike Alexander Fleming, the scientist who accidentally discovered penicillin in a buildup of mold, de la Fuente has pioneered methods that rely on AI algorithms to explore biological data, eventually paving the way for the discovery of antibiotics at a digital speed. His pursuit of this goal proved timely. Outside of his own research, technological advancements kept speeding along in the background which paved the way for his own progress. Machine learning and AI took off with the acceleration in computing power. And perhaps most consequential, the emergence of cheap genomic sequencing — a development that the New York Times recently hailed as generational:
Historians of science sometimes talk about new paradigms, or new modes of thought, that change our collective thinking about what is true or possible. But paradigms often evolve not just when new ideas displace existing ones, but when new tools allow us to do things — or to see things — that would have been impossible to consider earlier. The advent of commercial genome sequencing has recently, and credibly, been compared to the invention of the microscope.
In 2015, de la Fuente was recruited to be a postdoctoral fellow at the Massachusetts Institute of Technology, the mecca for AI research at the time, where he first began harnessing the convergence of these new breakthroughs. The initial feedback he received was not encouraging. “When I proposed that I wanted to use AI to design an antibiotic to treat infections, people looked at me like I was crazy,” de la Fuente recalls. “Nobody was really applying it to biology, and the general consensus was that it was impossible and not worth pursuing.” He recalls applying for faculty positions in numerous microbiology departments to apply his novel ideas, but not a single one called back.
Undeterred, at MIT, de la Fuente asked a fundamental question: Can computers create new antibiotics? With collaborators, he computationally designed an antibiotic that proved effective in preclinical mouse models. Intriguingly, the compound employed an innovative bacterial killing mechanism that was not originally programmed into the AI. This breakthrough showcased the ability of machines to generate emergent properties in biology, paving the way for new directions in antibiotic discovery.
In 2019, de la Fuente was recruited to Penn and posed another, related question that seemed to blur the lines between reality and science fiction: Can computers accelerate antibiotic discovery by mining the vast troves of biological data? This approach marked a significant departure from the conventional paradigm of unearthing antibiotics by laboriously sampling natural sources like soil. He predicted that, with the right algorithms, the world’s biological information could become a searchable repository for new antibiotics and other invaluable molecules. It was a bold jump into uncharted territory, reimagining how the digital realm might be harnessed to solve some of biology’s most pressing challenges.
“We can’t keep making one-step derivatives of the current drugs. We need creative strategies. Otherwise, we’re in real existential danger.” —Dr. Seth Shipman
Shortly thereafter, the de la Fuente Laboratory honed techniques for identifying and then synthesizing molecules based on genomic data — streamlining the process of discovery, by eliminating the need for organic samples. He started small, with the information about individual proteins. Then, the human proteome (the entire universe of proteins encoded in the human genome). And after the algorithms identified thousands of molecules which had previously not been studied for antibiotic purposes, de la Fuente kept going, searching beyond the flora and fauna of today’s world for tomorrow’s antibiotics.
“It was amazing to see that we could interrogate biology, biological data at the digital level. This led to the hypothesis that we would be able to find similar molecules … conserved all throughout evolution and across the Tree of Life,” he says. “We’ve found [potential] antibiotics in the wooly mammoth, ancient elephants, giant sloths, and many creatures from the past — as well as in living organisms, including humans and microbes.”
Since being recruited to Penn in 2019, de la Fuente has expanded that research. Molecular de-extinction is just one part of the lab’s work. The Neanderthal and “exctinctome” studies revealed whole new classes of peptides — fragments of protein — that might serve as a template for broader drug development, which the team has been synthesizing and testing in vitro and in vivo against pathogens. Altogether, the efforts of de la Fuente and his team have dramatically reduced the time required for the preclinical discovery phase of antibiotic research — from years to hours.
“Using AI and computers, we can discover hundreds of thousands of them in a typical day.”
The greatest medical advancement, redux
The first antibiotic used on humans was penicillin, discovered in Scotland in 1928. Though it’s arguably the most successful medicine in the history of the world, the influence of penicillin has not been all good. The success of the drug led to complacency among researchers. Even today, most of the antibiotics we consume are derivative of the techniques and molecules which produced penicillin. “For so long, there really didn’t need to be a lot of innovation, because it worked so well,” says Shipman.
Another factor in the decades-long decline in antibiotics research has been the failure of commercial companies to back their development. Antibiotics take longer to develop, cost more, and lack the promise of high sales — as they’re inherently short-term treatments, losing medical efficacy over time. Those tricky economics have contributed to a drying pipeline of research: In 1995, there were 586 papers published about antibiotics worldwide. In 2022, there were fewer than 200.
But that R&D equation could radically change for the better as a result of the Machine Biology Lab. “With traditional methods, it takes six or seven years to come up with one or two interesting compounds,” says de la Fuente, referring to preliminary discoveries that can lead to antibiotic drugs. “Now using AI and computers, we can discover hundreds of thousands of compounds in a typical day in the lab.”
The need for those discoveries is accelerating. Over the past century, society’s reliance on a fixed group of drugs has given rise to more resistance in the wild. Especially in the United States, antibiotics have been overprescribed for decades — not only in doctor’s offices, but also in agricultural fields and meat production. The Covid pandemic only accelerated that trend.
The result is a near future where things we take for granted, like a routine dental surgery or a seasonal bug, could be fatal. At this year’s World Economic Forum in Davos, scientists warned that due to AMR, the number of infectious deaths per year could rise to 10 million annually by 2050, if scientists don’t develop new drugs and tools to save us.
“As humans, we tend to invest only when the situation is extremely critical, like during the pandemic,” de la Fuente says. “Well, if we had been doing that for a while, then we would have probably prevented a lot of deaths”
The Machine Biology Lab is now home to a group of international researchers — ranging from undergraduates to post-docs, hailing from places including China, Italy, and India — who are working at the cross-section of natural and computer sciences. While their discoveries are years away from clinical development, de la Fuente is thinking about creating a company, and perhaps a nonprofit, that could bring those discoveries to market in time.
De la Fuente believes that a company is the best way to take the research through clinical trials and regulatory approvals, assuming it gets that far, but he’s not ruling out other avenues — like partnerships with nonprofits or licensing agreements with larger pharmaceutical companies. If the world ends up safer in the end, that’s all that matters.
“That’s really my biggest dream,” he says. “All the science that we’re doing using AI and these different tools, hopefully someday it helps improve the world by saving lives.”
